ARTICLES

PUBLISHED ONLINE: 12 JANUARY 2014 | DOI: 10.1038/NPHYS2842

Real-space tailoring of the electron–phonon

coupling in ultraclean nanotube mechanical
resonators
A. Benyamini1† , A. Hamo1† , S. Viola Kusminskiy2 , F. von Oppen2 and S. Ilani1 *

The coupling between electrons and phonons is at the heart of many fundamental phenomena in nature. Despite tremendous
advances in controlling electrons or phonons in engineered nanosystems, control over their coupling is still widely lacking.
Here we demonstrate the ability to fully tailor electron–phonon interactions using a new class of suspended carbon nanotube
devices, in which we can form highly tunable single and double quantum dots at arbitrary locations along a nanotube mechanical
resonator. We find that electron–phonon coupling can be turned on and off by controlling the position of a quantum dot along
the resonator. Using double quantum dots we structure the interactions in real space to couple specific electronic and phononic
modes. This tailored coupling allows measurement of the phonons’ spatial parity and imaging of their mode shapes. Finally, we
demonstrate coupling between phonons and internal electrons in an isolated system, decoupled from the random environment
of the electronic leads, a crucial step towards fully engineered quantum-coherent electron–phonon systems.

S
ome of the most well-known phenomena in molecular and
solid-state physics result from the coupling between electrons
and phonons. The resistivity of metals, ferroelectricity,
Peierls and Jahn–Teller instabilities and BCS superconductivity
are different facets of this coupling. In solids, electron–phonon
coupling is dictated by the lattice structure and the ensuing
electronic bands, leaving little room for tunability. Over the past
decades, tremendous advances have been made in the ability to
engineer materials on the nanoscale. On the electronic side, artificial
atoms—quantum dots—were created, providing extensive control
over their electronic spectrum1 which allowed the exploration
of a wide variety of phenomena inaccessible in bulk solids. On
the mechanical side, a growing variety of engineered systems2,3
enabled the study of mechanical phenomena on the nanoscale
and brought experiments closer towards controlling the quantum
state of mechanical resonators4–7 , as well as their coupling to
single spins8,9 , qubits5,10 and photons11 . These remarkable advances
contrast with the still-limited control over the coupling between
the electronic and mechanical subsystems, whose tailoring would
provide a remarkable toolbox for nano-electro-mechanical systems
in the classical and quantum regimes.
Carbon nanotubes constitute a particularly promising system
for tailoring the electron–phonon coupling. Their pristine lattice
recently enabled the realization of extremely clean electronic
systems12–14 , albeit still limited in length and complexity. Moreover,
their one-dimensional nature, light mass and large stiffness enabled
the creation of tunable mechanical resonators15,16 with high Q-
factors17,18 . Recent pioneering works19,20 coupled a carbon nanotube
resonator to a single quantum dot, demonstrating that the
mechanical frequency can be strongly affected by a single carrier
and that the correlated motion of electrons and vibrations can lead
to mechanical frequency softening21 .
In this work, we explore a new generation of suspended carbon
nanotube devices with wide-ranging local control, allowing the
formation and manipulation of single and double quantum dots
embedded into the mechanical oscillator. Using this unprecedented
level of control we engineer local field gradients that produce
controllable coupling between the electrons and the acoustic
transverse phonons. The advantage of this coupling over the
intrinsic strain-induced electron–phonon coupling is that it can be
tailored at will. Our measurements are done in the classical limit
(large vibration amplitudes) but because the interaction with the
external fields is linear in the carbon nanotube displacement, these
measurements probe the same electron–phonon coupling matrix
element that would be relevant in the quantum limit. We study the
dynamics of this coupling, and its spatial dependence, as well as
demonstrate mode-selective coupling in an isolated system.
The device (Fig. 1a) is created using the nano-assembly
technology we recently developed22 . It consists of a small-bandgap
carbon nanotube suspended between two metallic contacts above
five electrically independent gates. Above the metallic contacts the
carbon nanotube is hole-doped due to the contacts’ workfunction.
The suspended part, on the other hand, can be locally doped with
either electrons or holes by applying independent d.c. voltages to
the gates, Vg1 to Vg5 . A negative voltage on all gates dopes the entire
carbon nanotubes with holes, effectively creating a continuous
‘wire’ whose conductance is only weakly gate-dependent. However,
when a positive voltage is applied to one of the gates while
keeping negative voltages on the others, the nanotube’s segment
above this gate is doped by electrons, forming a pair of p–n
junctions that confine a quantum dot above this gate22 (Fig. 1b).
As we will show, not only does this dot act as a detector of the
local mechanical motion through its charge sensitivity, but more
importantly, it provides controlled local coupling between the
electronic and mechanical systems, which forms the fundamental
building block for this work.
The mechanical vibrations are measured through a standard
mixing technique15 . A radiofrequency signal with frequency f

1 Department of Condensed Matter Physics, Weizmann Institute of Science, Rehovot 76100, Israel, 2 Dahlem Center for Complex Quantum Systems and
Fachbereich Physik, Freie Universität Berlin, 14195 Berlin, Germany. † These authors contributed equally to this work. *e-mail: shahal.ilani@weizmann.ac.il

NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics 1

© 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLES NATURE PHYSICS DOI: 10.1038/NPHYS2842

a d ×10¬3

G (e2/h)
Nanotube 2 5e 6e 7e 8e 9e 10e

ce
0

ur

n
So

ai
Dr
87 Δf1

f (MHz)
4
2

5
3
1
te

te

te

te

te
My (a.u.)
Ga

Ga

Ga

Ga

Ga
83 ¬1 0 1

2.2 2.4 2.6 2.8 3.0

b E Egap
c
6 15 Vg3 (V)
e ×10¬3

dImix /df (pA MHz¬1)

x

G (e2/h)
Electron dot 1 1e 2e 3e 4e 5e 6e
Holes
Mx
¬My
Imix (pA)
0
215
0 0

f (MHz)
Vg1 Vg2 Vg3 Vg4 Vg5 My (a.u.)
z f +δf ¬1 0 1
¬dMx /df 209
f
¬4 ¬10
x 207.5 210.0 2.4 2.6 2.8 3.0 3.2 3.4 3.6
Imix(δ f ) = Mx + iMy f (MHz) Vg4 (V)

Figure 1 | A carbon nanotube mechanical resonator coupled to localized ultraclean quantum dots. a, Scanning electron micrograph of a device similar to
the one measured, with an 880-nm-long nanotube suspended 125 nm above five gates with a periodicity of 150 nm. Scale bar, 100 nm. b, Measurement
layout: d.c. gate voltages, Vg1 to Vg5 , locally dope the nanotube with electrons (red) or holes (blue). Mechanical motion is actuated by a radiofrequency
signal on gate 4 (frequency f) leading to a high-frequency modulation of the current, which is down-mixed to low frequencies using a weak probe signal of
frequency f + δf applied at the source. c, The mixing current, which is the current measured at frequency δf, has components that are in-phase (Mx ) and
out-of-phase (My ) with the drive; both are plotted as function of the drive frequency (blue and green, respectively). Also shown is the derivative dMx /df
(purple). d, Top: conductance, G, of a dot above gate 3 as a function of Vg3 . Bottom: corresponding mixing signal, My (colour map), measured for the first
mechanical mode, as a function of Vg3 and f. Dashed red line is a fit to a theory including only the static electron–phonon coupling, capturing the frequency
step across a Coulomb blockade peak. The dashed black line includes also the dynamical coupling (Supplementary Information 1). Their difference at the
centre of the Coulomb peak, 1f1 , gives the dynamic frequency softening. e, Similar measurement for the second mechanical mode with a dot above gate 4.
All measurements in this article are done at an electron temperature of T = 16 K as determined from the Coulomb peaks in the conductance.

is applied to an off-centre gate (gate 4) and a weaker ‘probe’ The first step in creating a tailored interaction between electrons
signal with frequency f + δf is applied to the source electrode and phonons is to study the underlying dynamics of their coupling,
(Fig. 1b). The former actuates the mechanical motion and mixes so far hardly explored experimentally owing to the limited control
down with the latter through the dependence of the current on over tunnelling barriers. Fundamentally, this coupling results from
the source–drain and gate voltages, ∂ 2 I /∂VSD ∂Vg , to produce a correlated mechanical and electronic motions: owing to carbon
low-frequency (δf ) current signal measured at the drain. When nanotube vibrations, electrons are pumped between the leads and
f is swept through a mechanical resonance, the carbon nanotube the quantum dot and their attraction to a biased gate causes a
vibration is enhanced, producing a sharp peak in the out-of-phase softening of the mechanical restoring force21 . This process involves
quadrature of the mixing signal, My , as well as in the derivative of an interesting competition between the vibrational frequency and
its in-phase quadrature with respect to f , dMx /df (Fig. 1c, more the electronic tunnelling rates, which we study here using a tunable-
details in Supplementary Information 1). Here, we use these peaks barrier quantum dot, formed in the resonator over the three central
interchangeably to trace out the mechanical resonance. gates and populated with holes (Fig. 2a). The quantum dot can
Figure 1d,e shows the measured gate dependence of the first be tuned across the entire range from a closed quantum dot to
two mechanical modes of the carbon nanotube resonator. The the open Fabry–Perot-like regime, with individual control over
motion is detected by a quantum dot formed on the resonator at the left and right tunnelling rates, ΓL and ΓR , by the side gates
a position of large movement (above gate 3 for the first mode and (1 and 5). We calibrate these rates using transport measurements
above gate 4 for the second mode, see illustrations). The bottom (Fig. 2b, details in Supplementary Information 2) and measure
panels show the measured My (colour map) as a function of the their independent effect on the mechanical softening. Starting with
voltage on the gate beneath the dot and of the drive frequency, the symmetric case, ΓL ≈ ΓR , we observe that the softening of the
f ; the top panels show the simultaneously measured conductance. first mechanical mode, 1f1 , drops with progressive pinching-off
For both modes the mixing signal is visible whenever the dot of the barriers (Fig. 2c I–III). Plotting the extracted 1f1 versus the
is conducting and, being proportional to the derivative of the total tunnelling rate ΓL + ΓR (Fig. 2d), we observe that the drop
conductance with respect to gate voltage, it is negative on one commences when the total tunnelling rate becomes comparable
side of the Coulomb peak (blue), positive on the other (red), and to the vibrational frequency, 2πf1 . This drop reflects the inability
zero at the peak (white). As in previous experiments, we observe of electrons to follow the mechanical motion and is reproduced
that for both modes, the resonance frequencies increase with gate by a theoretical calculation (dashed line, details in Supplementary
voltage owing to the tensioning of the resonator15 , that charging Information 5). Interestingly, even when only one barrier is open
by a single electron causes a discrete frequency jump19,20 and that (ΓR  2πf 1  ΓL ) the softening assumes the maximal value (Fig. 2c
the coupling between the electronic and mechanical degrees of IV). This demonstrates that contrary to the device conductivity,
freedom causes a sharp softening dip of the resonance frequency which is determined by the two barriers adding in series, the
concomitant with the Coulomb peak19,20 . This dynamical coupling softening is controlled by the two rates added in parallel, reflecting
will be used here to control the interactions between the mechanical that the relevant electrons can enter from either lead. Measurements
and electronic degrees of freedom. at very large tunnelling rates (Supplementary Information 3) show

2 NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics

© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE PHYSICS DOI: 10.1038/NPHYS2842 ARTICLES
a b Decreasing ΓL Log10(Gpeak[e2/h]) d 0.8
¬3.0 IV
II III
I ¬3.6
0.6
1.05 ¬4.2

Δf1 (MHz)
Decreasing Γ R
ΓL ΓR 1.00 0.4 I

Vg5 (V)
0.95 II
0.2

2πf1
0.90
Vg1 Vg5 IV 0.0
III 0 200 400 600 800 1,000 1,200 1,400
1.00 1.05 1.10 1.15 Γ L + Γ R (MHz)
Vg1 (V)

c e 7 ×10¬3
Decreasing ΓL ~ Γ R ΓL << Γ R
6 6

Q¬1

elec
I II III dMx /df (a.u.) IV dMx /df (a.u.)

Q¬1
1 1 5 3
89 87.6 0 88 0
88
f (MHz)

f (MHz)
4

elec
¬1 ¬1 I ¬2.00 ¬1.96

Q¬1
3 Vg3 (V)
88 87 86.6 87 2 II
¬2.04 ¬1.98 ¬1.98 ¬1.92 ¬1.82 ¬1.76 ¬1.92 ¬1.86 III

2πf1
1
Vg3 (V) Vg3 (V) Vg3 (V) Vg3 (V)
0
0 200 400 600 800 1,000 1,200 1,400
Γ L + Γ R (MHz)

Figure 2 | Dependence of dynamical electron–phonon coupling on the electron tunnelling rate. a, Schematics: a quantum dot of holes is created above
gates 2–4. Its left and right barrier tunnelling rates, ΓL and ΓR , are controlled by Vg1 and Vg5 . b, Measurement of the peak conductance at the Coulomb
blockade transition from 5 to 6 holes, Gpeak (colour map), as a function of Vg1 and Vg5 , from which ΓL and ΓR are independently extracted (see
Supplementary Information 2 for details). c, Mixing signal of the first mode, dMx /df (colour map), measured as a function of Vg3 and f for symmetric
tunnelling rates, ΓL ≈ ΓR (subpanels I–III), and for asymmetric rates, ΓL  ΓR (subpanel IV). d, Extracted softening, 1f1 , as a function of the total
tunnelling rate, Γ = ΓL + ΓR . The dashed line is a fit to the theory in Supplementary Information 6. The dotted vertical line marks the angular frequency of
the mechanical mode, 2πf1 . e, Inset: the inverse mechanical Q-factor, deduced from the frequency width of the resonance, plotted as a function of Vg3
across a Coulomb peak. Dashed line: theory fit (Supplementary Eq. 19). The difference between the on- and off- peak dissipation gives the electronic
contribution (Q−1 elec ). Main panel: extracted Qelec as a function of ΓL + ΓR for ΓL ≈ ΓR . Dashed line: theory fit (Supplementary Eq. 20). Error bars in d,e
−1

indicate 1 s.d. in 1f1 and in Q−1elec , respectively.

a similar reduction of 1f1 , this time due to gradual disappearance of effects, we ensure that all of the parameters that are relevant for
the Coulomb blockade phenomenon23 . In between these two drops, softening are similar for dots formed at the different locations.
there is a wide range of tunnelling rates for which the softening Specifically, all dots have the same number of electrons, similar
remains practically constant (Fig. 2d and Supplementary Fig. 4). In charging energies and similar gate couplings (Supplementary In-
this regime electrons enter sufficiently fast to establish electrostatic formation 4), and their tunnelling rates are chosen well within the
equilibrium at all times but not so fast as to broaden the Coulomb range where they do not affect the softening, as explained above.
blockade peaks beyond their thermal broadening. Interestingly, however, when measuring the softening of the first
The dependence on the total tunnelling rate is also remarkable phonon mode with dots at the five locations (Fig. 3a–e, illustration
for the dissipation of the phonon modes. As shown in the in each panel) we observe that it depends strongly on position: it is
main panel of Fig. 2e, the dissipation (inverse Q-factor) is non- weak near the contacts and increases continuously until reaching a
monotonous as a function of the electronic tunnelling rate, peaking maximum at the centre of the resonator. Figure 3f–j shows similar
when this rate is comparable to the vibration frequency. This might measurements for the second phonon mode. Again we observe
come as a surprise because, as in previous experiments19–21 , we find strong position dependence; however, in contrast to the first mode,
that the dissipation decreases monotonically as the gate voltage the softening is practically zero at the centre, and has its maxima
is detuned away from the Coulomb blockade peak, which is also above gates 2 and 4. Plotting the extracted softening for both modes,
associated with reduced electron tunnelling. Nevertheless, the peak 1f1 and 1f2 , as a function of the spatial position of the quantum dot
as a function of electron tunnelling is quantitatively described by (Fig. 3k,l), we find that they nicely follow the spatial displacement
our theory (dashed line, theory Supplementary Information 5) profile squared of the corresponding phonon modes (plotted as
and can be rationalized as a close analogue of the Debye peak lines). The coupling thus provides a direct imaging of the shapes
familiar from dielectric relaxation or viscoelasticity. Usually, the of the phononic modes in real space.
Debye peak is observed as a function of an indirect parameter The observed spatial dependence can be understood by
(temperature) that changes the rates in the system. Here, in calculating the local forces generated on the carbon nanotube
contrast, we observe it by directly controlling the relevant physical resonator due to single-electron charging of a localized quantum
parameter—the tunnelling rates of the electrons. The full control dot. In Supplementary Information 7 we show that this force can
that we achieve over the relative dynamics of the electrons and be described by an effective ‘electronic spring’ with a negative spring
vibrations will be used below to study previously unattainable constant, ‘attached’ at the position of the dot (illustrated in Fig. 3k).
regimes of their coupling. The spring constant does not depend on the position of the dot
A key feature that allows us to tailor the coupling between elec- along the tube, but the shift in the frequency of the combined
trons and phonons is the control over the real-space confinement system does: if the spring is connected at a node of the phononic
of the electrons. With five gates we can localize a quantum dot at mode it has no effect on its dynamics, whereas if it is connected
five different locations along the tube, and explore how its position at a location of large vibrational amplitude, it has a strong effect.
affects the coupling. To eliminate spurious position-dependence Indeed, perturbation theory shows (Supplementary Information 7)

NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics 3

© 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLES NATURE PHYSICS DOI: 10.1038/NPHYS2842

dMx /df (a.u.)

a b c d e 80 1 k First phononic mode
80 0
78.5 78.5 ¬1 0.0
77.5

Δf1 (MHz)

z12 (a.u.)
¬0.5 ¬⏐kelec⏐
f (MHz)

¬1.0
¬1.5
0.5 MHz

78 78 0 220 440 660 880

76.5 76.5
75.5
1.6 1.7 2.15 2.25 2.1 2.2 2.1 2.2 1.85 1.95 X (nm)
Vg1 (V) Vg2 (V) Vg3 (V) Vg4 (V) Vg5 (V)
dMx /df (a.u.)
f g h 211 i j 1 l
0 Second phononic mode
213 211
210 210 ¬1 0

Δf2 (MHz)

z22 (a.u.)
¬1
f (MHz)

¬2
0.5 MHz

¬3
209 207 207 0 220 440 660 880
206 206
1.36 1.46 2.18 2.28 2.85 3.05 2.2 2.3 1.6 1.7 X (nm)
Vg1 (V) Vg2 (V) Vg3 (V) Vg4 (V) Vg5 (V)

Figure 3 | Spatial dependence of the electron–phonon coupling, and direct imaging of the phonon modes. a–e, Mixing signals of the first phonon mode,
dMx /df (colour map), measured with quantum dots formed above each of the five gates (see illustrations). Vertical bars are 0.5 MHz. The softening shows
a clear dependence on the quantum-dot position. f–j, Similar measurements for the second mode, yielding a different position dependence. k,l, The
softening of the first and second modes, 1f1 and 1f2 , extracted from a–j and plotted as a function of the positions of the quantum dots, taken from the
lithographic positions of the gates. Lines: calculated amplitude squared of the corresponding phonon mode versus position in the beam (solid) and string
(dashed) limits (Supplementary Information 4). Inset to k: effective mechanical model: the force exerted on the nanotube by electrons flowing through a
localized dot is equivalent to a spring with a negative spring constant attached at the dot’s position. Error bars in k,l: x axis bars indicate uncertainty in the
location of the dot due to fabrication misalignment, and y axis bars indicate 1 s.d. in 1fi .

a 0 1 2 3 G (e2/h) b c d
×10¬3 First phonon mixing Second phonon mixing
1

My (a.u.)
3.0
0
2.8 2.8 2.8
¬1

2.5
Vg4 (V)

Vg4 (V)

Vg4 (V)

Vg4 (V)
) 2.7 2.7 2.7

2.0 2.6 2.6 2.6

2.0 2.5 3.0 2.0 2.6 2.7 2.5 2.6 2.7 2.5 2.6 2.7
Vg2 (V) Vg2 (V) Vg2 (V) Vg2 (V)

e f ¬dG/dV (a.u.) g dG/dε (a.u.)

1
0
First +δ V +δ V 2.8 2.8
¬1
Vg4 (V)

Vg4 (V)

2.7 2.7
Vleft

Second
¬δ V +δ V

2.6 2.6
V V
Vright
2.5 2.6 2.7 2.5 2.6 2.7
Vg2 (V) Vg2 (V)

Figure 4 | Tailored selective coupling between phononic and electronic modes in a double quantum dot. a, Conductance, G (colour map), of a double dot
defined above gates 2 and 4 measured as a function of Vg2 and Vg4 . The right, centre and left barriers are controlled by gates 1, 3 and 5, respectively. The
label (n,m) represents the number of electrons in each dot. b, Zoom-in on the (3, 4) to (4, 3) transition. The common mode and detuning gating directions
are labelled by V and ε. c, First-mode mixing, My (colour map), measured over the same voltage window as in b. d, The same for the second mode. In both
cases we subtract the electronic mixing signal measured away from the mechanical resonance and integrate over a small frequency window to project out
the frequency shifts due to softening (Supplementary Information 1). e, The mechanical motion of the different modes leads to different modes of effective
gating: for Vg2 = Vg4 = V, the first-mode vibrations lead to common-mode gating (blue arrow) and the second mode leads to detuning gating (green
arrow). f,g, Numerical derivatives of the conductance in b along the V and ε directions, respectively, showing good agreement with the measured
mechanical mixing (c,d).

4 NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics

© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE PHYSICS DOI: 10.1038/NPHYS2842
that the frequency shift is proportional to the amplitude squared
of the bare phonon mode at the location of the local spring, in
agreement with our observation. Beyond providing direct imaging
of the phonon modes24 , the above measurements demonstrate that
by moving the quantum dot in real space it is possible to turn on
and off its coupling to selected phonon modes, thereby creating
controllable coupling between these degrees of freedom.
We now take this coupled system to a new level of control, by
embedding for the first time a double quantum dot in a mechanical
carbon nanotube resonator. The added degrees of freedom allow
us to tailor selective couplings between specific electronic and
phononic modes, without the need to move the confinement
potential. The observed physics has interesting connections to
the coupling of electrons in double quantum dots and photons
reported recently25,26 . The double dot is formed above gates 2
and 4, which also act as the corresponding plunger gates, and
its left, right and centre tunnel barriers are controlled by the
remaining three gates. The measured double-dot conductance as
a function of the left and right plunger gate voltages (Fig. 4a)
shows an extremely clean charge stability diagram, down to
the single-electron limit. For the experiment we zoom onto a
symmetric charge transition vertex (Fig. 4b) and study the two
complementary facets of the electron–phonon coupling. We start
by exploring the effects of the phonons on the electrons, through
which we demonstrate the mode selectivity of this coupling. We
then demonstrate the complementary effect of internal double-dot
electronic modes on the phonons.
The coupling effects on the electrons are imprinted in the mag-
nitude and sign of the mixing signal, as opposed to the frequency
shift discussed so far. The former is isolated in Fig. 4c which plots
the mixing signal My , measured for the first mechanical mode
over the same voltage range as in Fig. 4b, but with the frequency
shifts integrated out (see caption). A similar measurement for the
second mechanical mode is shown in Fig. 4d. Curiously, we see
that the patterns of negative and positive mechanical mixing signal
(blue/red) are substantially different for these two modes. This
difference reveals the distinct way that different phononic modes act
on the electrons (Fig. 4e): in the first mode, the two dots are moving
in phase, together getting closer and further away from the plunger
gates. In the second mode, the dots move out-of-phase, with one
approaching and the other receding from the gates. This different
mechanical motion translates into different electrical gating: for
equal d.c. voltages on the plunger gates, the first mode gates the
double dot along the common-mode voltage direction (vector V in
Fig. 4b) whereas the second mode gates it along the detuning direc-
tion (vector ε in Fig. 4b). Each phonon mode can thus be mapped
onto an effective ‘gating vector’ in voltage space, and correspond-
ingly, its mixing signal should be the derivative of the conductance
along the direction of this vector. By taking the numerical derivative
of the conductance in Fig. 4b along the V and ε directions (Fig. 4f,g)
we indeed observe excellent agreement with the measured mixing
signals of the first and second modes, respectively. The above
measurement demonstrates two important aspects: first, it shows
that it is possible to use the electrons to directly probe the real-space
parity of the phonons. Furthermore, it demonstrates that each
phonon mode has a characteristic action on the electrons, captured
by its ‘gating vector’, providing a powerful tool for tailoring selective
coupling between these degrees of freedom.
So far we have studied the different aspects of tailored coupling;
however, this coupling was to random electrons tunnelling in and
out from the leads. An even more interesting case would be to
determine whether it is possible to couple the phonons to internal
electrons in an isolated system, thereby realizing a clean electron–
phonon coupled system insensitive to the random environment of
the leads. To check this we isolate the double dot from the leads
by symmetrically pinching-off its side barriers while maintaining a
NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics
© 2014 Macmillan Publishers Limited. All rights reserved.
ARTICLES
ε (V) ε (V)
¬3.07 ¬3.02 ¬2.85 ¬2.80
198.0 197.5
f (MHz)
f (MHz)
Mx (a.u.)
196.5 ¬1 0 1 196.0
Internal Internal and external
1.33
1.00
Δfi/Δfi,max
0.66
0.33
First mode
Second mode
0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2
γ = (ΓL + Γ R )/2πfi, (i = 1,2)
Figure 5 | Coupling of mechanical motion to internal electronic degree of
freedom in a double quantum dot, isolated from the leads. The first- and
second-mode softening is plotted as a function of their normalized
tunnelling rates γ = (ΓL + ΓR )/2πfi , (ΓL ≈ ΓR are the left and right barriers
tunnelling rates; fi is the mechanical resonance frequency of the ith mode).
The softening values are normalized by their asymptotic magnitude at large
γ , 1fi /1fimax . The double dot is gradually isolated from the leads by
symmetrically decreasing ΓL and ΓR and keeping fixed a
ΓC  2πfi (i = 1,2). The dashed line is a fit to the single-dot theory as in
Fig. 2, describing well the dependence of the first mode but not the
γ -independent softening of the second mode. Top insets: measured
softening for the points with the smallest (left) and largest (right)
tunnelling rate to the leads, together with a sketch of the relevant
second-mode electron tunnelling mechanisms in these two limits. Error
bars indicate 1 s.d. in 1fi /1fimax .
large internal tunnelling rate, ΓC  2πfi (i = 1,2). Figure 5 shows
the softening of the first and second modes measured in this regime
along their effective gating directions (V and ε respectively) and
plotted as a function of a normalized tunnelling rate to the leads,
γ = (Γ L + ΓR )/2πfi . As the double dot is disconnected from the
leads by lowering γ , we observe a clear quenching of the first-mode
softening, in complete analogy with the observation for the single-
dot case (Fig. 2). Remarkably, however, the detachment from the
leads leaves the softening of the second mode essentially unaffected
(evident by comparing the extreme cases γ  1, and γ > 1, (top
insets)). This intriguing observation can be understood by realizing
that a double dot has an internal electronic degree of freedom,
involving the transfer of charge between the dots, which provides
the correlated electron flow that induces the softening (illustration,
Fig. 5). This degree of freedom does not couple to the common-
mode gating of the first mode but directly couples to the detuning
gating of the second mode and thus softens only the latter, an effect
that is nicely captured theoretically (Supplementary Information
8). The above measurements thus clearly demonstrate strong and
selective coupling between phonons and internal electronic degrees
of freedom in an isolated system.
The observation of tailored coupling between internal electronic
and phononic degrees of freedom opens a wide range of new
possibilities. One example pertains to the coupling of phonons
to solid-state qubits, which use the singlet and triplet states of
5
ARTICLES
two electrons in a double quantum dot as their basis27,28 . Owing
to the Pauli blockade, an electron can shift between the dots
only in the singlet state but not in the triplet state, and thus
the second phonon frequency will dynamically couple only to the
former. This selective coupling thus provides a tantalizing new
route for transferring quantum entanglement from the electronic
to the mechanical subsystems, or even between distant qubits in
a multi-site lattice through a phonon ‘bus’29,30 . Generalizing the
physics demonstrated here in double dots to multi-site lattices that
are now well within reach22 would enable an even broader class
of experiments that explore bulk electron–phonon phenomena,
such as ferroelectricity, Peierls and Jahn–Teller instabilities, or
superconductivity, in an engineered nanoscale setting. Analogous
to the richness of quantum dot physics, made possible by the
extensive control over their electronic properties, the ability to
tailor the dynamics, spatial structure and selectivity of the coupling
between electrons and phonons, demonstrated here, will enable the
study of these phenomena in new regimes unattainable in bulk
systems, opening new frontiers for fundamental experiments in
condensed-matter physics.
Received 17 June 2013; accepted 13 November 2013;
published online 12 January 2014
References
1. Kastner, M. A. Artificial atoms. Phys. Today 46, 24–31 (1993).
2. Ekinci, K. L. & Roukes, M. L. Nanoelectromechanical systems. Rev. Sci. Instrum.
76, 061101 (2005).
3. Poot, M. & van der Zant, H. S. J. Mechanical systems in the quantum regime.
Phys. Rep. 511, 273–335 (2012).
4. LaHaye, M. D., Buu, O., Camarota, B. & Schwab, K. C. Approaching the
quantum limit of a nanomechanical resonator. Science 304, 74–77 (2004).
5. O’Connell, A. D. et al. Quantum ground state and single-phonon control of a
mechanical resonator. Nature 464, 697–703 (2010).
6. Teufel, J. D. et al. Sideband cooling of micromechanical motion to the quantum
ground state. Nature 475, 359–363 (2011).
7. Chan, J. et al. Laser cooling of a nanomechanical oscillator into its quantum
ground state. Nature 478, 89–92 (2011).
8. Arcizet, O. et al. A single nitrogen-vacancy defect coupled to a nanomechanical
oscillator. Nature Phys. 7, 879–883 (2011).
9. Kolkowitz, S. et al. Coherent sensing of a mechanical resonator with a
single-spin qubit. Science 335, 1603–1606 (2012).
10. LaHaye, M. D., Suh, J., Echternach, P. M., Schwab, K. C. & Roukes, M. L.
Nanomechanical measurements of a superconducting qubit. Nature 459,
960–964 (2009).
11. Aspelmeyer, M., Meystre, P. & Schwab, K. Quantum optomechanics.
Phys. Today 65, 29 (2012).
12. Cao, J., Wang, Q. & Dai, H. Electron transport in very clean, as-grown
suspended carbon nanotubes. Nature Mater. 4, 745–749 (2005).
13. Kuemmeth, F., Ilani, S., Ralph, D. C. & McEuen, P. L. Coupling of spin and
orbital motion of electrons in carbon nanotubes. Nature 452, 448–452 (2008).
14. Steele, G. A., Gotz, G. & Kouwenhoven, L. P. Tunable few-electron double
quantum dots and Klein tunnelling in ultraclean carbon nanotubes. Nature
Nanotech. 4, 363–367 (2009).
6 NATURE PHYSICS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturephysics
© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE PHYSICS DOI: 10.1038/NPHYS2842
15. Sazonova, V. et al. A tunable carbon nanotube electromechanical oscillator.
Nature 431, 284–287 (2004).
16. Leturcq, R. et al. Franck–Condon blockade in suspended carbon nanotube
quantum dots. Nature Phys. 5, 327–331 (2009).
17. Hüttel, A. K. et al. Carbon nanotubes as ultrahigh quality factor mechanical
resonators. Nano Lett. 9, 2547–2552 (2009).
18. Ganzhorn, M. & Wernsdorfer, W. Dynamics and dissipation induced by
single-electron tunneling in carbon nanotube nanoelectromechanical systems.
Phys. Rev. Lett. 108, 175502 (2012).
19. Lassagne, B. et al. Coupling mechanics to charge transport in carbon nanotube
mechanical resonators. Science 325, 1107–1110 (2009).
20. Steele, G. A. et al. Strong coupling between single-electron tunnelling and
nanomechanical motion. Science 325, 1103–1107 (2009).
21. Woodside, M. T. & McEuen, P. L. Scanned probe imaging of single-electron
charge states in nanotube quantum dots. Science 296, 1098–1101 (2002).
22. Waissman, J. et al. Realization of pristine and locally tunable one-dimensional
electron systems in carbon nanotubes. Nature Nanotech. 8, 569–574 (2013).
23. Meerwaldt, H. B. et al. Probing the charge of a quantum dot with a
nanomechanical resonator. Phys. Rev. B 86, 115454 (2012).
24. Garcia-Sanchez, D. et al. Mechanical detection of carbon nanotube resonator
vibrations. Phys. Rev. Lett. 99, 085501 (2007).
25. Frey, T. et al. Dipole coupling of a double quantum dot to a microwave
resonator. Phys. Rev. Lett. 108, 046807 (2012).
26. Petersson, K. D. et al. Circuit quantum electrodynamics with a spin qubit.
Nature 490, 380–383 (2012).
27. Hanson, R., Kouwenhoven, L. P., Petta, J. R., Tarucha, S. & Vandersypen,
L. M. K. Spins in few-electron quantum dots. Rev. Mod. Phys. 79,
1217–1265 (2007).
28. Pei, F., Laird, E. A., Steele, G. A. & Kouwenhoven, L. P. Valley-spin blockade
and spin resonance in carbon nanotubes. Nature Nanotech. 7, 630–4 (2012).
29. Cirac, J. & Zoller, P. Quantum computations with cold trapped ions. Phys. Rev.
Lett. 74, 4091–4094 (1995).
30. Turchette, Q. et al. Deterministic entanglement of two trapped ions. Phys. Rev.
Lett. 81, 3631–3634 (1998).
Acknowledgements
We acknowledge O. Auslaender, E. Berg, F. Kuemmeth, P. L. McEuen, A. Shnirman and
A. Yacoby for useful discussions and comments on the manuscript, and D. Mahalu for
the electron-beam writing. S.I. acknowledges the financial support by the Legacy
Heritage Foundation (ISF, No. 1267/12), the Bi-National Science Foundation (BSF, No.
710647-03), the Minerva foundation (No. 780054), the ERC starters grant
(QUANT-DES-CNT, No. 258753), the Marie Curie People grant (IRG, No. 239322),
and the Alon fellowship. S.I. is incumbent of the Z. William and E. B. Novick career
development chair. F.v.O. acknowledges support through Schwerpunktprogramm SPP
1459, SFB 658 and a Helmholtz Virtual Institute.
Author contributions
A.B., A.H. and S.I. performed the experiments, analysed the data and contributed to its
theoretical interpretation. S.V.K. and F.v.O. developed the theoretical model.
Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to S.I.
Competing financial interests
The authors declare no competing financial interests.