Development of levitated electromechanics of nanodiamond in a Paul trap Anas Almuqhim A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy of University College London. Department of Physics and Astronomy University College London December 18, 2019
157
Embed
Development of levitated electromechanics of nanodiamond ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Development of levitatedelectromechanics of nanodiamond in
a Paul trap
Anas Almuqhim
A dissertation submitted in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
of
University College London.
Department of Physics and Astronomy
University College London
December 18, 2019
2
I, Anas Almuqhim, confirm that the work presented in this thesis is my own.
Where information has been derived from other sources, I confirm that this has been
indicated in the work.
Abstract
This thesis outlines the development of an experimental platform to explore recent
theoretical proposals to create macroscopic spatial quantum superposition using a
levitated nanodiamond containing nitrogen vacancy centres (NV). The work has
demonstrated a method for electrodynamic levitation of nanodiamond and explored
the feasibility and limitations of this system for experiments in macroscopic quan-
tum mechanics.
A range of electrical trap geometries were explored to determine their suitability
for diamond levitation. A linear quadrupole trap was chosen and two different traps
were designed, constructed and tested at atmospheric pressure and under vacuum.
One trap was of a conventional linear Paul trap design, which was integrated with
a microwave antenna as one of the electrodes for excitation of NV centres in the
nanodiamond. A more cost effective trap was also designed and constructed from a
printed circuit board. This design was easy to fabricate and had a larger numerical
aperture for enhancing signal detection. Although not eventually used in this work
it has found application in other levitation experiments in the laboratory.
Most of the work in this thesis utilises a conventional linear Paul trap with inte-
grated microwave excitation. The magnetic field strength and the energy density
of the microwave field within the trap was modelled and were found to be suitable
for excitation of NV centres. Using this trap we demonstrated optically detected
magnetic resonance (ODMR) of the NV centres of diamond placed in the trap. NV
fluorescence from microdiamond in this system was used to investigate the depen-
dence of NV photoluminesence as a function of temperature, laser power and gas
pressure. It was found that the temperature change not only affected the resonance
Abstract 4
frequency but also the ODMR contrast. The contrast reached its peak at about of
11.7± 0.2 % at 380± 35 K down to 3.2± 0.2 % at 657± 20 K. We demonstrated
the ability to trap 100 nm diamond down to 4×10−3 mbar and observed the NV
photoluminesence at atmospheric pressure.
Impact statement
The principle of superposition is a central part of quantum theory that has been
demonstrated in the microscopic world for particles ranging from electrons all the
way up to very massive molecules containing thousands of atoms. However, it is
still an open question as to whether it is a universal principle that holds for all mass
scales. An answer to this question for even larger systems, such as nanoparticles,
will not only have impact on our knowledge of the universality of quantum me-
chanics but also on the limits of macroscopic quantum technologies that utilise this
principle.
The work outlined in this thesis represents the development of an experimental plat-
form, using levitated nanodiamond, that eventually aims to test the macroscopic
limits of quantum mechanics using matter-wave interferometry. This work has de-
signed, tested and demonstrated methods for electrodynamic levitation of nanodia-
mond, as well as the detection of embedded nitrogen vacancy centres (NVs), both
of which are central to this type of matter-wave interferometry. The trapping of
nanodiamond in vacuum within a Paul trap, may have an impact on the way future
high-mass, matter-wave interferometry experiments can be carried out. In addition,
the work in this thesis has highlighted some significant technical problems, such as
that the internal heating of nanodiamond, must be overcome before nanodiamond
matter-wave interferometry can be realised.
Acknowledgements
First, all praises to Allah for the strength and his blessing in completing this thesis.
I would like also to express my sincere gratitude to my supervisor Professor Peter
Barker for his exceptional support and for patiently offering me guidance and advice
on my research with practical issues that go beyond the textbooks. Without him, this
thesis would not have been possible.
I am also grateful to King Abdulaziz City for Science and Technology (KACST)
for funding my studies at UCL.
Many thanks also to the past and present members of the Barker group, for their
exceptional support and for providing me with an excellent atmosphere to carry out
my research. It would have been a lonely road without them.
I would like to give special thanks to Dr. Sultan Ben Jaber for granting me access
to Raman spectroscopy in the chemistry department at UCL. I would also like to
thank Dr. Khaled Aljaloud for giving me access to CST software in the electrical
engineering department at UCL. I must also thank Dr. Masfer Alkahtani for the NV
diamond samples.
Finally, my deepest gratitude goes to my beloved parents and brothers. They are
always supporting and encouraging me with their best wishes.
7.5 ZFS frequency and ODMR contrast obtained from Fig 7.15. . . . . 115
7.6 ZFS frequency and ODMR contrast obtained from Fig 7.17. . . . . 117
7.7 ZFS frequency and ODMR contrast data from Fig 7.19. . . . . . . . 118
7.8 Coupling parameters between temperature and laser intensity. . . . . 121
Chapter 1
Introduction
1.1 Background
Testing the limits of quantum mechanics in the macroscopic domain is currently
of significant interest in physics. More specifically, is quantum mechanics limited
by scale? In its formalism, conventional quantum mechanics does not show any
limitation on mass. The linearity of the Schrodinger equation shows that quantum
superposition states should exist at all scales. This is true in the microscopic regime
as demonstrated for light by Thomas Young in 1801 in his famous double slit ex-
periment and for electrons by Davisson and Germer [1]. However, whether the
superposition principle holds for much larger masses is an open question.
A range of theories have been posited for the rapid collapse of the wavefunction
which would prevent the creation of a superposition on macroscopic mass and
length scales. These include the model by Ghirardi-Rimini-Weber (GRW) [2] which
modifies the Schrodinger equation by introducing nonlinearity and stochastic terms
that scale with mass and size. The Continuous Spontaneous Localization (CSL)
model, which is derived from this work, has been used to predict collapse rates and
has been compared with existing microscopic and macroscopic experiments [3].
Although these models predict collapse, they are phenomenological, and therefore
do not provide a physical mechanism for the process of collapse. However, it is
possible that gravitational interaction with a superposition could lead to wavefunc-
tion collapse and localisation, and this has been initially explored by Penrose and
18
Diosi [4, 5]. Experimental technology is now bringing these theories into the realm
of testability, in which conventional decoherence processes, which would also lead
to collapse, are minimised or controlled. Recent proposals include the use of matter-
wave interferometry [6–10] and optomechanical techniques [11–16].
In 1999, Arndt et al. [17] reported remarkable results proving the de Broglie hypoth-
esis for large objects using C60 fullerene. The particle with highest mass reported
to date in a quantum superposition consists of 104 amu and is an organofluorine
molecule [18, 19]. Since then, however, experimental technology for matter-wave
interferometry has been the barrier to performing these type of experiments for
larger objects. For example, decoherence such as black body emission and absorp-
tion of thermal photons, scattering of light from optical gratings, and collisions with
residual gas molecules, are sources of decoherence that must be controlled. When
silicon particles of 107 amu are diffracted by an optical grating, the scattering of
photons from the grating leads to a reduction in fringe visibility by approximately
10 % [19]. Of experimental importance, it is challenging to have a monochromatic
source of particles which have a narrow enough mass and velocity range such that
the interference pattern is not smeared out through interference with a wide range
of de Broglie wavelengths. Finally, as the particles have such a small de Broglie
wavelength the interference pattern is increasingly difficult to resolve with conven-
tional detectors [18–20].
A different approach to matter-wave interferometry of macroscopic particles was
proposed by Bose et al. [21]. This is a type of Ramsey interferometry that uses
levitated mesoscopic nanodiamond containing singlet spin (S = 1) nitrogen vacancy
centres (NV). This scheme is attractive because it may offer a way of increasing
the mass range up to 109 amu. In this approach a spatial superposition is created
within a trap by using a single NV spin embedded in a larger nanodiamond parti-
cle. In principle this can be a single spin embedded in a macroscopic particle, but
in the interests of concreteness we will adopt the proposal as described in refer-
ences [21, 22]. In this proposal, the atomic spin S = 1 of the NV centre is coupled
to the centre-of-mass motion of the nanodiamond using a magnetic field gradient.
19
An outline of the experiment, and how the interferometer works, is described in
Fig 1.1.
B
B
MW pulse
B
B
t
0 +1 −1
z-axis
mg
Figure 1.1: Time evolution of the experimental set up proposed by Bose et al. [21]. Ananodiamond levitated in an optical trap under strong magnetic field gradientoriented along the z axis. To begin with, the NV diamond centre is initial-ized |Ψ(0)〉 = |β 〉|0〉, and then a microwave (MW) pulse is applied to createa superposition state with equal probability |0〉 → 1√
2(|+ 1〉+ | − 1〉), which
oscillate in opposite directions. By tilting the experiment, a gravitational phaseshift (due to a difference in height) will occur between the two spin states|Ψspin〉 = 1√
2(|+ 1〉+ ei∆φ | − 1〉), which will be measured at the end of the
interferometer. At the end, another MW pulse is sent to measure the spin1√2(|+1〉+ |−1〉)→ |0〉 and perform the Ramsey measurement.
Here a single NV centre inside the nanodiamond is levitated in an optical trap un-
der ultrahigh vacuum to isolate it from the environment. The diamond’s centre-
of-mass motion is in a coherent state |β 〉 and the internal spin state is |Sz〉 = 0
such that the wavefunction is initially governed by |Ψ(0)〉= |β 〉|0〉. At time t = 0,
and under a strong magnetic field gradient oriented in the z-direction, a microwave
pulse is applied to create a spin superposition state of |+ 1〉 and |− 1〉 with equal
probability, |0〉 → 1√2(|+ 1〉+ | − 1〉). The individual spins states of the superpo-
sition are displaced by the field gradient and each oscillates in different harmonic
potentials. At t = T (T equal to the period of oscillation), and by tilting the ex-
periment, a gravitational phase shift (due to a difference in height) occurs between
the two spin states. This introduces a relative phase, due to gravity, between the
spin states |Ψspin〉 = 1√2(|+ 1〉+ ei∆φ | − 1〉), where the phase shift ∆φ = −12λ∆λ
h2ωzT
,
∆λ = 12mgcos(θ)
√h
2mωzand λ = 3µ0mzz0
4π|z0|5gNVµB
√h
2mωz. Here, ωz is the trapping
20
frequency in the z axis, mz is the dipole moment of the static magnetic field along
the z-axis generated by a magnet located at z0, and µ0 is the permeability of vac-
uum. µB denotes the Bohr magneton, gNV the Lande factor and h is the reduced
Planck constant. Finally, another MW pulse is used to measure the probability of
the |+1〉 and |−1〉 to return to the |0〉 state, such that 1√2(|+1〉+ei∆φ |−1〉)→ |0〉.
This experiment has advantages over conventional interferometry, in that it does
not require ground state cooling - although it does require some cooling to retain
the linear regime of the harmonic potential well [21–24]. Additionally, it can, in
principle, use the same trapped particle for many trials of the same experiment. In
this scheme, the spatial separation is only ∼ 1 pm for 100 nm particle using a mag-
netic field of ∼ 1× 104 T/m. This is, therefore, a weak superposition, though it is
enough to verify the creation of a superposition.
A follow up work from Bose et al. [22] demonstrated the ability to create larger
spatial separations up to 100 nm (the size of the particle), by allowing the particle
to free fall. This process is described in Fig 1.2. At time t0 and under a strong mag-
netic field gradient oriented along the z-axis, the optical trap is turned off. A free
falling particle under gravity is in a spin state |β 〉 (|+1〉+|−1〉)√2
. The two spin states
accelerate in opposite directions under the inhomogeneous magnetic field (as in the
Stern Gerlach effect). At t1, a MW pulse is applied to flip the spin, causing the
states to decelerate then accelerate towards each other. At t2, another MW pulse is
applied to slow the spin states so the two components overlap at t3. At t3, a third
MW pulse is sent to close the interferometer; an overall phase which affects one
side of the interferometry arms more than the other will result in a modulation that
indicates the up and down of the interference, as above.
Comparing the two previous schemes proposed by Bose et al., the first scheme has
the advantage of repeating the interference since the particle stays always in the
harmonic well of the trap. However, the free falling scheme offers larger superpo-
sition. It requires an initial feedback cooling and the superposition state is detected
by repeated measurements.
21
𝑩
mg
z-axis
z1
+a
t
+a
-a
-a
-a -a
+a
+a
t0
t1
t2
t3 Detector
𝛽 +1 + −1
2
+1 −1 MW pulse
Figure 1.2: Time evolution of the experimental set up proposed by Bose et al [22] by drop-ping the particle. At t0, the optical trap is turned off, and the particle falls undergravity which is in spin state β
(|+1〉+|−1〉)√2
. The two spin states accelerate in op-posite directions under the inhomogeneous magnetic field. At t1, a MW pulseflips the spin, causing the two spin states start to decelerate, then come to a turn-ing point where they accelerate again. At t2, another MW pulse is used to slowthe two components before they overlap at t3. At t3, a third MW pulse is used toclose the interferometer and carry out the Ramsey interference measurement.
22
1.2 Motivation
To carry out these proposals experimentally, many experimental difficulties must
be addressed. These include the requirement to levitate the nanodiamond under
vacuum to prevent environmental decoherence. In addition, we must be able to ma-
nipulate the spin and then measure the spin state at the end of the protocol. The spin
coherence lifetime must be considerably longer than the trap frequency. A typical
value for NV centres in nanodiamond is 100 microseconds and milliseconds for mi-
crodiamond [25–27]. This implies the trap frequencies must be well in excess of 1
kHz.
There is currently considerable research being undertaken to investigate the viability
of levitating NV diamond particles for matter-wave interferometry [23, 24, 28, 29].
For example, levitation in vacuum by an optical tweezer using 1550 nm or 1064 nm
laser beam has been demonstrated [23, 24, 29]. A 532 nm green laser was used to
initialise the spin, but this, as well as the trapping laser, unfortunately causes heat-
ing of the levitated NV diamond. The resulting increased temperatures increases
the probability of a non-radiative transition. This reduces the effectiveness of the
initial spin polarization required to start the interferometry, and also reduces the
fluorescence from the NV centre which is used to evidence the spatial superposi-
tion [23, 29]. To overcome this problem, the optical trap can be replaced by an
electrical one.
Benson et al. [30] demonstrated the trapping of micron-sized NV diamond in a
linear Paul trap under moderate vacuum conditions, and observed NV fluorescence.
Delord et al. [31] reported the observation of NV fluorescence from 10 µm diamond
at 2× 10−2 mbar. However, the 532 nm excitation laser at 700 µW was powerful
enough to heat up the particle which led to trap loss. This is due the impurities of
the NV diamond sample used. This could be addressed by using purer samples.
This PhD thesis outlines the development of a new experimental platform to explore
the aforementioned recent theoretical proposals [21, 22] to create macroscopic spa-
tial superposition using levitated nanodiamond, containing NV.
23
1.3 Thesis structureThe structure of this thesis is as follows:
• Chapter 2: An outline of the properties of the nitrogen vacancy centre in
diamond, relevant for its use in matter-wave interferometry.
• Chapter 3: A description of the trapping principles of Paul traps, and design
considerations for trapping nanoparticles.
• Chapter 4: The design and modelling of trap geometries suitable for nanopar-
ticle levitation.
• Chapter 5: A description of the fabrication, loading and characterisation of
Paul traps. This includes the production and evaluation of two of the traps
outlined in chapter 4.
• Chapter 6: The design and modelling of in-trap microwave excitation of NV
centres.
• Chapter 7: A study of the photoluminescence of NV diamond from micron-
sized diamond as a function of temperature, laser intensity and gas pressure.
• Chapter 8: A description of levitated nanodiamond experiments in a Paul trap.
• Chapter 9: Conclusion and future research.
Chapter 2
The nitrogen vacancy centre in diamond
relevant to matter-wave interferometry
2.1 Optical and spin propertiesThe nitrogen vacancy centre (NV) in diamond has remarkable properties, such as a
robust spin coherence lifetime and the ability to perform electron spin manipulation
at room temperature. The defect centre of a diamond lattice consists of a nitrogen
atom bonded to three carbon atoms and connected to a local vacancy, as shown in
Fig 2.1.
N C
CC
VC
CC
Figure 2.1: The NV centre in diamond. This consists of one nitrogen atom and three carbonatoms surrounding a central vacancy, trapped in the diamond crystal lattice.
The ground state of the NV has a singlet electron spin (S=1), with spin states
ms = -1, 0, 1 as shown in Fig 2.2. The ms = ±1 levels are offset from the ground
state by 2.87 GHz without an applied magnetic field because of spin-spin interac-
25
tions [32]. This is known as Zero Field Splitting (ZFS). In the |3E,ms = 0〉 excited
state these levels are offset by 1.42 GHz [33].
+ 1
− 1
0
+ 1
− 1
0
MW2.87 GHz
≈ 630-800 nm ISCGreen Light
510 - 540 nm
Ground state
Exited state
Orbital Spin
Excite
High probability
Low
probability
3𝐸
3𝐴2
Red
Fluorescence
637 nm
+ 1
− 1
2𝛾𝐵
Magnetic field
Figure 2.2: The electronic level structure of the NV centre in diamond. The |3A2〉 groundstate and the |3E〉 excited state has three spin states: ms = -1, 0, 1. The tran-sition between these two states is spin conserving. The |3A2,ms = ±1〉 statesare offset from the |3A2,ms = 0〉 state by D = 2.87 GHz, known as Zero FieldSplitting (ZFS). The excited state is populated from the ground state by opticalexcitation using green light (wavelength between 510-540 nm). Electrons inthe |3E,ms = 0〉 excited state decays to the |3A2,ms = 0〉 ground state throughtwo channels. A higher probability transition is observed as 637 nm red fluo-rescence, and lower probability transition via intersystem crossing (ISC). Forelectrons in the |3E,ms = ±1〉 state, decay occurs with higher probability tothe |3A2,ms = 0〉 ground state through the ISC, and with lower probability tothe |3A2,ms =±1〉 ground states. The emission spectrum is approximately be-tween 630 and 800 nm.
Electronic transitions from the |3A2,ms = 0〉 ground state to the |3E,ms = 0〉 ex-
cited state occur by optical excitation using light in the green part of the spectrum,
from 510 nm - 540 nm [34]. The electron then decays to the ground state via two
channels from the |3E,ms = 0〉 state.
The first, fast transition, with higher probability, occurs through radiative decay
with an energy of 1.945 eV. This is observed as red light at 637 nm, and is known
as the Zero Phonon Line (ZPL) of the NV− state, which is the intrinsic energy
difference between the lowest phonon energy level of the |3E〉 and |3A2〉 states.
26
The second, lower probability, transition is through a non-radiative process (also
known as a dark state) via intersystem crossing (ISC) [33,35]. This state was found
to be the neutral charge state (NV0) by Waldherr et al. [36], and this was verified
by Aslam et al. [34]. The NV0 state has a photoluminescence spectrum centred at
610 nm and a ZPL at 575 nm.
A 532 nm laser is commonly used for optical excitation in experiments.
The photoluminescence spectrum is usually acquired between 630 nm to
800 nm to cover the vibrational side bands that extend from the ZPL at
638 nm of the NV− state [33]. This spectrum is normally centred at 680 nm as
shown in Fig 2.3.
Figure 2.3: Fluorescence spectra of single NV centre in diamond showing the 532 nm op-tical excitation laser, the ZPL of NV0 at 575 nm and NV− at 638 nm with theirphonon side bands, adapted from reference [33].
2.2 Electron spin measurement
State preparation of the spin for the proposed matter-wave interferometry experi-
ments requires manipulation of the ground state spin states of the NV centre. The
most straight-forward approach to this is to perform what is commonly known as
electron spin resonance (ESR) measurements (also called electron paramagnetic
resonance EPR) [37]. This is done by tuning a microwave field across the ESR res-
onance, which, in diamond, is at 2.87 GHz for the NV− colour centre in diamond.
27
This will lead to a transition from the |3A2,ms = 0〉 state to the |3A2,ms =±1〉 state.
This transition is observed as a reduction of the photoluminescence count and as a
dip in the recorded ESR spectrum, as shown in Fig 2.4. This process is known as
optically detected magnetic resonance (ODMR).
By applying an external magnetic field (B), the spin ms = ±1 degeneracy will be
lifted by the Zeeman effect with a splitting of 2γB, where γ , the electron gyro-
magnetic ratio, is 2.8 GHz/T. This splitting is detected as two dips in the ODMR
spectrum as shown in Fig 2.4. The frequency separation between the peaks can be
correlated to the strength of the applied magnetic field [33, 35].
Figure 2.4: ESR spectrum of NV− under different external magnetic field (B) strengths.The frequency separation between the peaks can be correlated to the strengthof the applied magnetic field, adapted from reference [38].
In general, NV centres can be found in four different orientations [111], [111], [111]
and [111] within the nanodiamond crystal (see Fig 2.5) which are observed as 8 dips
in an ODMR spectrum [39–42]. The spectral separation between ms =±1 states is
a function of the alignment of the magnetic field with the NV orientations [39–42].
For one NV centre, we expect to observe two dips if the magnetic field is parallel to
the NV axis. For an ensemble, we expect to observe from 2 to 8 dips, depending on
the orientation of NV’s within the sample and the their alignment with the magnetic
28
field [39–42]. For example, Fukui et al. [39] reported the observation of two dips
in the ODMR spectrum for an ensemble, arguing that the magnetic field is parallel
to the [111] direction, and the 10 NV’s within the sample have a [111] orientation.
In a matter-wave interferometry experiment, the NV spin axes should be confined
during the preparation of the spin superposition state, and any rotational motion of
the trapped particle in the axial direction of the ion trap must be avoided. That could
be done by depositing a magnetic material on the NV diamond and use magnetic
field to align the particle [43].
NV V
NN
V
N
V
y
x
z
𝟏 𝟏 𝟏 𝟏 𝟏 𝟏 𝟏 𝟏 𝟏 𝟏 𝟏𝟏
Figure 2.5: NV centre orientations. NV axis can be found parallel to [111], [111], [111] or[111].
2.3 Creating a spin superposition
A superposition of the |ms = ±1〉, required for matter-wave interferometry as de-
scribed in chapter 1, can be created in the presence of a magnetic field by applying
two microwave pulses. Initially, green light is used to spin polarise the sample such
that the NV centre is in the |ms = 0〉 state. Following this, a microwave field pro-
vides a resonant π/2 pulse such that 50% of the population in the |ms = 0〉 state
is transferred to the |ms = −1〉 state. Another microwave π pulse is applied that
promotes the remaining population of the |ms = 0〉 state to the |ms =+1〉 state.
For single NV centre there now exists a superposition of | < ψ|ms = −1 > |2 =
|< ψ|ms =+1 > |2 = 12 . The relative phase of this superposition can be set by the
relative phases of the microwave fields [44].
29
2.4 Spin coherence lifetime
Although we have described how the spin superposition can be created, it is im-
portant to understand how the environment effects experiments with NV centres.
Particularly important is the spin coherence lifetime as this sets the timeframe over
which any experiment can be undertaken. Here we have two channels of spin relax-
ation: spin-lattice relaxation time T1 (also called longitudinal relaxation), which is
the transition time between |ms = 0〉, and |ms = ±1〉 and spin-spin relaxation time
T2 (also called transverse relaxation), which is the spin decoherence time [33,45]. In
bulk diamond, T1 is in the range of milliseconds at room temperature, and seconds
up to a few minutes at low temperature [46,47], while T2 is of the order of microsec-
onds at room temperature [25]. In nanodiamond, the value of T1 is in the range of
several microseconds, while T2 is in the range of hundreds of nanoseconds [45].
These values can be increased by decreasing the concentration of embedded NV
centres to prevent any spin-spin interaction between NV’s [48]. The lifetime for the
spatial separation of the ms =±1 superposition states is governed by the T2 time.
2.5 Temperature and pressure dependence
An important consideration for any experiment with NV centres is how temperature
affects the spectroscopy and the spin coherence lifetime. It is possible to use the
zero field splitting (ZFS); the NV centre spin resonance frequencies; for thermom-
etry [33, 49]. As described in section 2.1, the ZFS is the energy difference between
the ms =±1 state and the ms = 0 state, and it is given by D=2.87 GHz for the ground
state, where D is denoted as the ZFS parameter. Firstly, the temperature changes
affect the ODMR spectrum. This is seen as a shift in the ZFS and a reduction in
the PL intensity. This shift in ZFS is explained by local thermal expansion [49,50].
This explanation has been questioned, as it was limited to studies of just the ther-
mal expansion effect [35], though Doherty et al. [51] have since found that the shift
is associated not only with thermal expansion but also with electron-phonon inter-
actions. However, this explanation also depends on impurities and the size of the
diamond sample. A complete mechanism for this shift remains an open question.
30
Here we will focus on three reported results that are commonly used to define tem-
perature dependence of the ZFS [49, 50, 52]. These results could be categorised in
three temperature regions based on the shift in the ZFS parameter, D. These are
defined as low, middle and high temperature regions.
In the low temperature region from 5.6 K to 295 K, Chen et al. [50] reported the
temperature change in the ZFS parameter denoted as D by a fifth-degree polynomial
the shift in D due to the local strain, which ranges from 2 to 9 MHz [37] and
∆pressure = 1.5 Hz/mbar is the linear shift in D with pressure [53]. Note that,
reducing the pressure from atmospheric pressure to 2.2× 10−4 mbar leads to a
slight temperature rise from room temperature (about 11 mK), which is negligi-
ble in comparison with the temperature change associated with the total shift in
D and the uncertainty in equation 2.2 which is about ±20 K. In this region, the
shift in D is about -56 MHz (from 2.87 GHz down to 2.814 GHz) with gradient
dD/dT =80 kHz/K at 300 K up to dD/dT =170 kHz/K at 700 K.
31
From this we can see that the ZFS parameter D shifts up from 2.87 GHz when the
temperatures decreases, and vice versa when the temperature increases. Increas-
ing the temperature will also increase the probability of a non-radiative transition
through the singlet state, which is observed as a reduction of the PL intensity and
ODMR contrast [49,50,52]. The resulting increase in temperatures not only reduces
the effectiveness of the initial spin polarization required to start the interferometry
but also reduces the fluorescence from the NV centre, which is used to evidence the
spatial superposition. For example, levitation by optical tweezers using a 1550 nm
or 1064 nm laser beam in vacuum is problematic, mainly as a result of the trapping
laser [23, 24, 29]. Optical traps produce an increase in temperature, due to absorp-
tion of the trapping laser light, and these temperature changes affect the contrast of
the electron spin resonance (ESR) spectrum, the intensity of fluorescence emission
and the sharpness of the ZPL peak, especially when the laser power is high [23,29].
In other words, lower laser power leads to a clearer ESR spectrum and higher pho-
toluminescence emission. Moreover, the background gas acts as a heatsink, so, at
lower pressure, the internal temperature of the levitated beads increases and subse-
quently the nano-beads burn at between 10 mbar and 5 mbar [23,54], depending on
the trapping laser wavelength. To address this issue, the optical trap can be replaced
with an electrical one, and this approach is examined in this thesis.
2.6 ConclusionThis chapter described the properties of the NV centre in diamond which are im-
portant for the proposed matter-wave interferometry experiment. These properties
mean that the experiment will require an optical field to initialise the spin, a mi-
crowave antenna to manipulate the spin, and an optically detected magnetic reso-
nance (ODMR) system to readout the spin. A microwave antenna will be integrated
into a Paul trap, and this will be explored in chapter 6. The ODMR detection system
for the trap will described in chapter 7 and 8. Of particular importance is the shift
in zero field splitting (ZFS), which can be used to reveal the temperature changes
induced by a laser in the trap. This will be described in chapter 7.
Chapter 3
Paul traps for levitation
3.1 Trapping principles
Since Wolfgang Paul introduced ion traps in 1953 [55], they have become a useful
tool for many applications requiring mass selection, or for capturing charged parti-
cles in an isolated environment [56]. There are many different types, with the most
common having cylindrical, hyperbolic and linear electrode geometries. All use an
oscillating electric field that leads to the confinement of a charged particle in all
three spatial dimensions [57].
In order to understand the dynamics of a charged particle inside the Paul trap, we
will start by describing the force that governs the motion of a charged particle inside
the potential of a quadrupole field.
The force on a charged particle is given by [57]:
~F =−Q5φ , (3.1)
where Q is the charge and φ is an electric potential in three dimensional space given
by:
φ = G(Ax2 +By2 +Cz2). (3.2)
Here, G is the field gradient and A, B and C are constants which depend on the
geometry. As the Laplacian52φ = 0, this implies that A + B + C = 0. For example,
by considering a two dimensional quadrupole field where A = -B = 1, then C = 0.
33
The potential φ in equation 3.2 is then given by:
φ = G(x2− y2) =φ0
2r02 (x
2− y2), (3.3)
where φ0 is the applied potential (either AC or DC), and r0 is the distance from
the trap centre to the electrode surface as shown in Fig 3.1. This potential can by
realised by four hyperbolic electrodes with infinite length in the z dimension. How-
ever, cylindrical rods are commonly used for greater optical access and simplicity
in fabrication, as described later in this chapter.
Substituting φ from equation 3.3 in equation 3.1, the force can be written as:
~F =−2Qφ0
2r02 (x x− y y). (3.4)
-
++ +
-
F
F
FF
r0
E
x
y
Figure 3.1: A diagram that shows the electrode configuration of a two dimensionalquadrupole field. The potential inside the field is given by equation 3.3. Apositively charged particle near the trap centre will experience an attractiveforce toward the trap centre in the x direction, and a repulsive force from thetrap centre in the y direction.
Consider a positively charged particle near the trap centre, governed by equation
3.4. This particle will be attracted towards the trap centre in the x direction (the
positive x term in equation 3.4) and repulsed from the trap centre in the y direction
(the negative y term in equation 3.4). The opposite occurs for negatively charge
34
particles. This motion implies that it is impossible to confine a charged particle
in all spatial dimensions using a static electric field. This is known as Earnshaw’s
theorem [58]. However, an oscillating (AC) electric field can be used to obtain a net
confinement in the x and y directions toward the trap centre. Consider the electrodes
shown in Fig 3.2 where the potential is given by:
φ0 = 2(U +V cos(Ωt)). (3.5)
The voltages applied to the electrodes U and V are DC and AC respectively, and Ω
is the angular frequency of the oscillating potential.
Figure 3.2: A diagram that shows a four electrode structure of the Paul trap. The fourelectrodes are connected to an AC potential 2(U +V cos(Ωt)). The top andbottom electrodes are connected to the +(U +V cos(Ωt)) terminal and the rightand left electrodes are connected to the −(U +V cos(Ωt)) terminal.
In order to understand how the AC potential leads to a net confinement of a levitated
particle in the x− y plane, we visualise the electrostatic fields and the potential
using a SIMION simulation program. This program solves the Laplace equation
for a particular electrode geometry using a finite difference method [59, 60]. The
two dimensional AC potential forms an oscillating hyperbolic saddle surface which
is flipped from up to down over a half cycle as shown in Fig 3.3, a and Fig 3.3,
c. Consider a positively charged particle near the centre of the trap. At a certain
time t = 0, this particle will be repulsed from the electrodes with a positive voltage
and attracted toward the negative electrodes (see Fig 3.1 and Fig 3.3, a), but after a
35
half of the oscillation period at t = T/2, the electrode voltage has changed sign and
will now attract the particle as shown in Fig 3.3,c. By using appropriate values for
the potentials U and V with a particular oscillation time, the average force pushes
the particle towards the centre of the trap and the particle will be confined in both
x and y directions [57]. The quadratic potential is produced by hyperbolic shaped
equipotential lines as shown in Fig 3.3, b and Fig 3.3, d.
x
y
Potential
+
(a) The saddle-point potential during the firsthalf of the AC cycle at time t = 0.
x
y
Potential
(b) The equipotential lines at t = 0.
x
yPotential
+
(c) The saddle-point potential after half of theAC cycle at time t = T/2.
x
yPotential
(d) The equipotential lines at t = T/2.
Figure 3.3: The hyperbolic saddle surface generated by AC voltage over a half cycle. Theblue ball represents a positively charged particle. The AC potential forms anoscillating hyperbolic saddle surface. In the first half of the AC cycle, theparticle is confined in the x electrode direction (the positive curvature) and anti-confined in the y direction (the negative curvature). After half of the AC cycle,the saddle surface has flipped and the particle is confined in the y direction andanti-confined in the x direction. The particle will be confined in both x and ydirections at a particular AC frequency, where the flipping rate of the saddlesurface is faster than the time required for the particle to escape. This diagramhas been generated using SIMION, where the positive electrodes are held at+300 V and the negative electrodes at -300 V.
36
3.2 Linear quadrupole trap
The linear Paul trap is an example of the two dimensional quadrupole field described
in the previous section, but with additional DC electrodes for confinement in the
remaining axis. This trap consists of four long rod electrodes: two of them with
an applied AC potential, while the other two are at ground (the AC potential could
also be connected between the opposite pair of electrodes). At each end of these
electrodes are two DC electrodes known as endcaps as shown in Fig 3.4, where z0
is the distance between the endcaps and the trap centre, r0 is the distance from the
trap centre to the electrode surface.
x
y
z
Figure 3.4: A conventional linear quadrupole trap. The trap consists of four parallel rods,two connected to an AC potential, and the other two at ground potential. A setof additional DC electrodes at each end of the AC electrodes are used to formthe endcaps.
The potential in the quadrupole trap can be expressed in terms of the geometrical
parameters of the electrodes and the voltage applied to them [57, 61–63]. This is
given by:
φ =
[V2
cosΩt(
1+x2− y2
r20
)]+
[Uz2
0
(z2− 1
2(x2 + y2))] , (3.6)
where V is an AC voltage applied to the AC electrodes, U is a DC voltage applied
37
to the endcaps, z0 is the distance between the endcaps and the trap centre and, r0
the distance from the trap centre to the electrode surface. This potential can be seen
as the sum of the AC potential (first term) and the endcap potentials (second term).
By putting this potential into equation (3.1), we obtain the equations of motion of a
particle of charge Q and mass m in the quadrupole trap:
x =−Qm
(Vr2
0cosΩt−U
z20
)x, (3.7)
y =Qm
(Vr2
0cosΩt +
Uz2
0
)y, (3.8)
z =−2QUmz2
0z. (3.9)
These can be put into the form of Mathieu equations:
d2udζ 2 +(au−2qu cos(2ζ ))u = 0, (3.10)
where u represent x, y or z, and
ax = ay =−12
az =−4QU
mΩ2z20, (3.11)
qx =−qy =−2QV
mΩ2r20,qz = 0, (3.12)
ζ =Ωt2. (3.13)
These equations have a standard solution within the region of stable trapping for
parameters a and q, and that satisfy the condition |a| << q2 < 1 for the first sta-
bility region as shown in Fig 3.5, b. It should be noted that, in the second stability
region, denoted as regions B and C in Fig 3.5, a, where the value of a and q are
high, the micromotion can be of the same order as the macromotion [64]. Here, a
higher voltage and driving frequency for trapping are required. This motion will be
discussed in the next section.
38
A
B
C
Stability regions
0 1 2 3 4 5 6-4
-2
0
2
4
qu
au
(a) Mathieu stability diagram in two dimensions. The overlap-ping regions between x and y solutions represent the stability re-gions of the trap. The region denoted as A is the first stabilityregion (also known as the lowest stability region) while regionsB and C are the second stability regions.
q = 0.908u
First stability region
0.0 0.2 0.4 0.6 0.8 1.0-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
qu
au
(b) The first stability region (A) near the origin of the Mathieustability diagram. This region represents the commonly operatingregion for the linear Paul trap.
Figure 3.5: Stability diagram for the linear Paul trap in two dimensions. This diagramshows the numerical solution of Mathieu equations. The blue regions representthe x-stable solutions of Mathieu’s equations while the red regions represent they-stable solutions. The diagram shows a symmetry around the au axis. Duringthe experiment, only the positive part of the au axis is considered, though thenegative part could be achieved by swapping the polarity of the trap shown inFig 3.2.
3.3 Secular motion and micromotionThe motion of a charged particle in the trap can be described by Mathieu equa-
tions. The solution of these equation shows a harmonic oscillation with two types
39
of motion: a macromotion (also called a secular motion or trapped motion) and a
micromotion. In order to understand these two motions, we can look at an approxi-
mate solution to Mathieu equations under the condition a,q << 1 [57, 61] where:
u(t) = u0 cos(ωut + τu)(
1+qu
2cosΩt
). (3.14)
The variables u0 and τu determine the initial amplitude and phase respectively, ob-
tained from the initial particle position and velocity, and where the secular motional
frequency is given by:
ωu =Ω
2
√au +
q2u
2. (3.15)
This harmonic motion of a charged particle at frequency ωu and amplitude u0 is
called the macromotion or secular motion. The fast drive frequency Ω applied to
the trap induces motion at that frequency, but with a lower amplitude (related to
the cosΩt term). This motion is superimposed on the secular motion and called the
micromotion. Due to the large contrast between the frequencies of the slow macro-
motin and the fast micromotion (under the condition a,q << 1), the motions can
be seen as two separate motions. However, although this is true in the first stability
region, in higher stability regions the amplitude and frequency of the micromotion
can be of the same order as the secular motion [64]. The charged particle motion
gives the average kinetic energy over one period of the macromotion [61]:
EKu =12
m⟨u2⟩∼= 1
4mu2
0
(ω
2u +
18
q2uΩ
2)
∼=14
mu20ω
2u
(1+
q2u
q2u +2au
)∼=
14
mu20ω
2u︸ ︷︷ ︸
macromotion
+14
mu20ω
2u
(q2
uq2
u +2au
)︸ ︷︷ ︸
micromotion
.
(3.16)
In the z direction, where qz = 0, the micromotion term in equation 3.16 vanishes((q2
zq2
z+2az
)≈ 0)
and then the total kinetic energy can be described by the macro-
40
motion term. That means the micromotion can be minimized by keeping the particle
near the trap centre (close to the AC null point). In the x and y directions, and un-
der the stability condition |au|<< q2u < 1, the kinetic energy of the micromotion is
equivalent to the kinetic energy of the macromotion, since the expression(
q2u
q2u+2au
)in the micromotion term in equation 3.16 must be almost equal to 1 in order to
satisfy the stability condition. The trapped particle cannot be positioned at the trap
centre at all times, as it will still fluctuate around the saddle point and the micromo-
tion will remain. Assuming a secular motion which is in thermal equilibrium with
a bath of temperature Tx,y,z. From the equipartition of energy, the average value of
the kinetic energy can be written as:
⟨EKz
⟩=
14
m⟨u2
tz
⟩ω
2z =
12
kBTz, (3.17)⟨EKx,y
⟩=
12
m⟨
u2tx,y
⟩ω
2x,y = kBTx,y, (3.18)
where T is the particle temperature, kB is Boltzmann’s constant and⟨u2
t⟩
is the
variance amplitude of the thermal motion. From equation 3.17 and 3.18, we can
see that the energy of the macromotion can be reduced be decreasing⟨u2
t⟩, which
could be achieved by cooling [61, 65].
3.4 Pseudopotential approximationThe AC potential in equation (3.6) could be simplified to a time-independent poten-
tial by using a pseudopotential. This is basically the time averaged potential over
one oscillation cycle (2π/Ω), and is of the form:
ϕpse =14
QmΩ2 E2 =
14
QmΩ2
(|∇ΦAC(x,y,z)|2
). (3.19)
Now, by using equation (3.19), we can rewrite the potential in equation (3.6) as:
φ =
[14
QV 2
mΩ2r40
(x2 + y2)]+[QU
mz20
(z2− 1
2(x2 + y2))] (3.20)
41
Substituting again into equation (3.1) to obtain the equations of motion, we get:
x =−(
Q2V 2
2m2Ω2r40− QU
mz20
)x, (3.21)
y =−(
Q2V 2
2m2Ω2r40− QU
mz20
)y, (3.22)
z =−(
2QUmz2
0
)z. (3.23)
It can be seen from the equation in the z direction, where only the DC voltage is
applied, that the secular frequency is equal to:
ωz =
√2QUmz2
0. (3.24)
In the x and y directions, the secular frequency is given by:
ωx,y =
√Q2V 2
2m2Ω2r40− QU
mz20. (3.25)
It should be noted that, in none of the previous equations, did we account for the
effect of gravity, which shifts the particle from the trap centre. This can be compen-
sated by using a DC offset voltage [66].
3.5 Geometrical efficiency factorsThe equations in the previous sections assume hyperbolic shaped electrodes which
produce a quadratic potential. For non-hyperbolic electrodes, geometrical efficiency
factors are used to account for the non-quadratic potentials which allow the use of
Mathieu equations. Considering a linear Paul trap with cylindrical rods, where the
effect of non-hyperbolic shape electrodes can be written as [67, 68]:
Φ =
[14
QV 2η2
mΩ2r40
(x2 + y2 +σzz2)]+[κQU
mz20
(−εx2 +(1− ε)y2 + z2)] . (3.26)
42
This potential yields secular frequencies given by [68]:
ωx =
√Q2V 2η2
2m2Ω2r40− ε
2κQUmz2
0, (3.27)
ωy =
√Q2V 2η2
2m2Ω2r40− (1− ε)
2κQUmz2
0, (3.28)
ωz =
√Q2V 2η2σz
2m2Ω2r40+
2κQUmz2
0, (3.29)
where η , σz, κ and ε are the geometry efficiency factors [68]. The role of each of
these factors is as follows. The factor η takes into account the effective AC potential
of the electrode shape to the quadratic AC potential in the radial direction, where
( η ≤ 1 ). In the case of an ideal AC potential, η = 1. The factor κ has the same
interpretation as η but for the DC potential in the z direction. The factor σz takes
into account the effect of the residual potential along the z direction from the ap-
plied potential in the x and y directions. This could be negligible, since σz 1 [68].
This factor is equal to zero in the ideal case. The last geometry factor is ε , which
is an anisotropy factor that describes the effect of anisotropy in the DC potential in
the radial direction. The value of the factor ε is equal to 0.5 in the radial symmetric
case [68].
The hyperbolic shaped equipotential lines required for the Paul trap can be tech-
nically achieved using hyperbolic shaped electrodes. Commonly, however, cylin-
drical rods are used, which allow good optical access and simplify the fabrication
process. To get equipotential surfaces that closely approximate hyperbolic elec-
trodes requires that the radius of the electrode be equal to 1.1468 r0 [69] as shown
in Fig 3.6. More recent studies have shown that this number should be between
1.12 r0 to 1.13 r0, where r0 is the distance from the trap centre to the electrode
surface [70–72].
43
(a) Hyperbolic electrodes, r0 = 50 mm. (b) Cylindrical electrodes, r0 = 50 mm.
(c) The potential between the two x electrodes, where black dots rep-resent the potential for the hyperbolic electrodes and red dots representthe potential for the cylindrical electrodes.
Figure 3.6: Quadrupole equipotential lines for hyperbolic and cylindrical electrodes. Theapplied voltage is 300 V in the opposing x electrodes while the y electrodes aregrounded. The cylindrical electrode diameter is 57 mm. The potential alongthe dashed line in both (a) and (b) is shown in (c). The potential distribution forthe cylindrical electrodes is close to that of the hyperbolic electrode, since theelectrode radius is chosen to be 1.1468 r0. This comparison is modelled usingSIMION.
44
3.6 ConclusionThis chapter is a theoretical description of the dynamics and stability of charged
particles in Paul traps for levitation. This background will be used in the following
chapters to design and examine different trap geometries that are suitable for NV
diamond. For levitated NV diamond, it is important that the traps are as deep as
possible, with a large opening angle between the trap electrodes to collect maxi-
mum light. The quadratic field of an ideal trap suggests a hyperbolic electrode trap
design. However, most traps, including those that described in subsequent chapters,
are not hyperpolic in shape. The geometrical efficiency factors introduced in this
chapter will allow us to take into account the effect of different electrode geometries
introduced in the next chapter. The pseudopotential approximation introduced here
will be used to define and evaluate the depth of these different traps. We can also
use the formalism developed in this chapter to determine the stability criteria of a
particle levitated in such traps.
Chapter 4
Paul trap design
4.1 OverviewThe appropriate trap for our experiment should satisfy two main criteria. Firstly,
the potential well should be as deep as possible to make trapping easier, and to
enable it to hold the particles for long periods. Secondly, a large opening angle (or
numerical aperture) between the trap electrodes is required for efficient detection
of the fluorescence signal from the NV nanodiamond levitated in the trap. Some
established designs that meet these criteria were evaluated and these are described
in the following sections.
4.2 Linear quadrupole trap with endcapsAs described in section 3.2, the linear quadrupole trap consists of four parallel
electrodes that carry AC voltage, as well endcap electrodes with static voltages,
as shown in Fig 3.4. The geometry of this design was chosen to be as close as
possible to the hyperbolic electrode trap. This was achieved by using cylindrical
electrodes with radius R≈ 1.1468r0 and with infinite length. Since the real trap has
a finite length and has endcaps in the z direction, the length of the AC electrodes
was chosen to be twice the gap between the opposing electrodes (4r0). This reduces
the effect of the residual potential along the z direction from the applied potential
in the x and y directions. Considering the length of the AC electrodes, the endcap
electrodes are spaced apart by z0 ≥ 4r0 from the trap centre, with no restrictions on
length and size. In this trap design, shown in Fig 4.1 and Fig 4.2, r0 = 1.75 mm, and
46
each AC electrode has a length of 7 mm (= 4r0) and a radius of 2 mm (= 1.146r0).
The endcaps (DC electrodes) have a 10 mm length and a 2.5 mm in radius. This
trap enables optical access from two directions without any additional modification.
Light is detected through the AC electrodes, which limits the numerical aperture of
this trap to N.A.≈ 0.24 in one direction, as shown in Fig 4.2. Although this is not
a particularly high N.A., we cannot make it higher because we want to maintain
the ideal ratio of the distance between the trap centre and electrode surface to the
electrode radius (≈1.1468), to produce a quadrupole potential close to the ideal hy-
perbolic electrode.
The potential along the x coordinate (the y coordinate should be similar to x due
to radial symmetry) was first investigated using COMSOL Multiphysics. This soft-
ware solves the Laplace equation for a particular electrode geometry using finite
element methods. The result was then compared with SIMION for validation, both
software packages give the same results as seen in Fig 4.3.
x (mm)
y (mm)
z (mm)
Figure 4.1: Linear quadrupole with endcaps. The AC and ground (GND) electrodes havea length of 7 mm and a diameter of 4 mm. The distance between the ACelectrodes is 3.5 mm. The DC electrode length is 10 mm and the diameter5 mm. The AC and DC electrodes are separated by 2 mm.
47
1.75 mm1.75 mm
4 mm
3.7
5 m
m
450
103.807°
31.193°
27.614°
31
.19
3°
NA= 0.24
Figure 4.2: Opening angle of the linear quadrupole trap. The four circles represent the ACelectrodes. The two opposing pair electrodes are separated by 3.5 mm. Theradius of each electrode is 2 mm. The numerical aperture (NA) is about 0.24.
Figure 4.3: Comparison of the calculated potential between the two AC electrodes in thelinear quadrupole with endcaps trap. The applied voltage is 300 V and the sep-aration between AC electrodes is 3.5 mm. The red dots represents the potentialcalculated using COMSOL, the black dots represents the potential calculatedusing SIMION.
48
The electric field distribution has been modelled using COMSOL Multiphysics. The
computed electric field, where both the DC and AC voltage (applied to two of the
AC electrodes; the other two are grounded) is held at 300 V is illustrated in Fig 4.4,
a and Fig 4.4, b, where the colours shows the electric field strength in V/m. The
pseudopotential approximation, which is the time-averaged potential at the secular
frequency, is given by:
ϕpse =14
QmΩ2 E2 =CE2, (4.1)
where C = 14
QmΩ2 , Q and m are the particle charge and mass respectively, Ω is the
driving frequency and E is the electric field strength. This equation shows that the
potential, as seen by any charged particle, is proportional to E2. Here we charac-
terise the trap by the value of E2. Fig 4.4, c is a plot of E2 along the x axis, and
Fig 4.4, d along the z axis. The maximum E2 of this trap is about 3.4×1010 V2/m2
along the x and y axes, and about 4.1×109 V2/m2 along the z axis.
For the matter-wave interferometry experiment, we considered the lowest depth of
the three directions, which represent the energy barrier that a particle must over-
come to escape the trap. We assumed a diamond with a diameter of 100 nm and
one elementary charge, and a driving frequency of 1 kHz. In this case, the depth of
the trap is about 2.3 eV which is equivalent to T= 5.3×104 K.
Two further simulations were carried out in order to calculate the geometry effi-
ciency factors. in the first of these, the AC electrodes were set at 1 volt, and the
endcap DC electrodes grounded, to calculate η and σz. In the second simulation,
the DC electrodes were set at 1 volt, and the AC electrodes grounded, to obtain ε
and κ . These parameters are listed in table 4.1, where we can see that ε is equal
to 0.5 due to the symmetry in the radial coordinate and that the value agree with
that reported by Madsen et al. [68]. Also, the value of σz is 3.24×10−3, so we can
neglect the z2 term in equation 3.20, which was also reported by Madsen et al. [68].
It should be pointed out that η = 0.5 because the voltage is applied at two of the
AC electrodes, whereas the other two were grounded. In case of a voltage applied
to all AC electrodes, this number will be doubled. This value is almost equal to the
49
ideal case (where η=1), because we maintain the distance from the trap centre to
the electrode surface at (≈ 1.1468r0).
(a) Electric field contours in the x-y plane.Colours is used to indicate the electric fieldstrength in V/m. Dark blue circles represent theAC electrodes. The electric field is at a minimumin the trap centre, at point (0,0).
(b) Electric field contours in the z-x plane. Thetwo dark blue rectangles in the centre representthe AC electrodes, while the others are the end-caps. The electric field is at a minimum along theAC electrodes, where particles could be trapped.The field by endcaps will prevent the particlesfrom escaping the trap in the z direction.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 0
1
2
3
4
0
1
2
3
E2 x1010
(V/m
)2
x ( m m )(c) E2 along x axis (E2 along y is similar to x).The maximum E2 is about 3.4×1010 V2/m2.
z ( m m )(d) E2 along z coordinate. The maximum E2 isabout 4.1×109 V2/m2. Note that the scale here isdifferent to (a), to recognise the plot shape.
Figure 4.4: Calculated electric field for the quadrupole trap with endcaps. The AC elec-trodes are held at 300 V and the endcaps at 300 V. The separation between theAC electrodes is 3.5 mm and the distance between endcaps in z direction is11 mm.
Table 4.1: Geometry efficiency factors for the linear quadrupole trap with endcaps. For theideal case η = 1, σz = 0, ε = 0.5 and κ = 1.
η σz ε κ
0.5 3.24×10−3 0.5 0.255
50
Overall, this standard design affords a deep trap depth of about 2.3 eV
(T= 5.3× 104 K), and is compact in size, while the opening angle between elec-
trodes is moderate to assure a purely quadratic potential. In addition, as Benson
et al. [30] observed NV fluorescence from microdiamond, using a linear Paul trap
with a N.A. less than 0.2, the N.A. for this trap is sufficient for NV fluorescence
detection.
4.3 Quadrupole trap on printed circuit boardA second trap design was developed and built to improve optical access as com-
pared to the conventional linear quadrupole trap discussed so far. This was made by
deforming the conventional hyperbolic electrodes into a set of planar electrodes on
a Printed Circuit Board (PCB), as shown in Fig 4.5 a and Fig 4.5, b. This approach
was implemented recently in an ion trap [73,74]. The deformation reduced the size
and increased the optical access to the trapped particle.
We first developed a design based on the conventional linear quadrupole trap. This
consisted of four parallel rectangular AC electrodes, each with a length of 4.3 mm
and a width of 2 mm as shown in Fig 4.6, c. A set of four perpendicular rectangular
DC electrodes, with the same dimensions as the AC electrodes, and separated by
2z0 (≈ 5.8 mm), were used for axial confinement. The numerical aperture, when
viewed from the top of the circuit board, is approximately 0.68, as shown in Fig 4.5,
d. This is a significant improvement for fluorescence detection, when compared
with the linear quadrupole trap. However, there is a lack of optical access in the y
direction, as it is blocked by the PCB substrate. This limits the accessibility in this
direction as shown in Fig 4.5, a.
Following the same procedure as that used for the linear quadrupole trap with
endcaps, the electric field is calculated for AC and DC voltages of 300 V, as
shown in Fig 4.6, a and Fig 4.6, b. The maximum E2 for this design is about
3.51× 1010 V2/m2 in the y direction, 1.4× 1010 V2/m2 in the z direction and
1× 1010 V2/m2 in the x direction as shown in Fig 4.6, c and Fig 4.6, d. It can be
seen that, in the radial (y-z) plane, the depth in the y direction is less than that in the
51
z direction. This is because the PCB trap is not as radially symmetric as the linear
quadrupole trap, as the electrodes are much closer to each other in the z direction
(see Fig 4.5). Comparing this trap with the linear quadrupole trap, we find that,
between the AC electrodes in the radial plane, the PCB trap is slightly deeper by
3.2% in one direction and shallower by 59% in the other direction. In terms of the
distance between the endcaps (2z0), the PCB trap is deeper by 144% compared to
the linear quadrupole trap, because the distance between the PCB endcaps is shorter
by 5.2 mm. It is worth mentioning that the PCB trap is only deeper because the
dimensions are smaller compared to the linear quadrupole trap. If the PCB trap had
similar dimensions to the linear quadrupole trap, the potential would be less deep
in all directions. For the matter-wave experiment, the trap depth is about 5.5 ev
(T=13× 104 K) for a charge-to-mass ratio of 0.1 C/kg and a drive frequency of
1 kHz. This means it is deeper by 60.5%, compared to the linear quadrupole trap.
The geometry efficiency factors for this trap are listed in table 4.2. These factors
shows the deviation from the hyperbolic trap. The factor η = 0.35 shows that the
trap depth is reduced by 15 % compared to the ideal case for AC voltage. Here, the
voltage is applied at two of the AC electrodes, whereas the other two are grounded.
In the case of voltage applied to all AC electrodes, this number will be doubled.
This number will be doubled. The factor κ = 0.349 has the same interpretation as
η but for the endcap potential which is reduced by 62 %. It can be seen that the
factor ε is equal to 1.31 due to the asymmetry in the radial (y-z) plane (in the case
of radial symmetry ε = 0.5 as for the linear quadrupole with endcaps). Except for
κ , these factors compare unfavourably to the linear quadrupole trap. However, this
will not affect the trapping principles discussed in the previous chapter, although
the trap depth will be shallower.
Table 4.2: Geometry efficiency factors for a PCB trap [75]. In the ideal case η = 1, σz = 0,ε = 0.5 and κ = 1.
η σz ε κ
0.349 6.2×10−3 1.31 0.382
52
xy
z
(a) A 3D model of the PCB trap. The size of thePCB is 25 mm × 25 mm with a thickness of1.6 mm. Each electrode is labelled with theapplied voltage. The trap includes a slot be-tween the electrodes where the particles shouldbe trapped.
(b) Layout of the PCB trap electrodes without thesubstrate, to show the top and bottom electrodes.This trap consists of four parallel rectangular elec-trodes - two of them with an applied AC potential,and the other two at ground. At each end of theseelectrodes are two perpendicular rectangular DCelectrodes.
Hole (5.8×1.6mm)
Copper (4.3×2mm)
GND
Copper (4.3×2mm)
AC
Top View
x
y
Co
pp
er (
2×4
.3m
m)
DC
Co
pp
er (
2×4
.3m
m)
DC
(c) Top view.The AC and the ground (GND) electrodes have a lengthof 4.3 mm and a width of 2 mm . The DC electrodes have the samedimensions as the AC and ground electrodes. The slot between theelectrodes has a length of 5.8 mm and a width of 1.6 mm. The AC andDC electrodes are separated by 0.75 mm.
PCB substrate (1.6mm)
Copper (70µm) GND
Copper (70µm) AC
Copper (70µm) AC
Copper (70µm) GND
Side View870µm
0.8mm 0.8mm47.4°
85.2°
N.A. = n*sin θ = 0.68
y
z
(d) Side view. Each electrode has a thickness of 70 µm while the thickness of the substrate is1.6 mm. The numerical aperture is about 0.68.
Figure 4.5: Geometry of the PCB trap. The four-rod geometry of the linear quadrupoletrap is changed to a set of plane electrodes to improve optical accessibility. Theelectrodes are printed on a dual sided PCB.
53
(a) Electric field contours in the y-z plane.Colours indicate the electric field strength inV/m. Dark blue lines represent AC and ground(GND) electrodes. The upper right and lower leftelectrodes are AC, while the upper left and thelower right are GND. The electric field is at aminimum in the trap centre, at point (0,0).
(b) Electric field contours in the x-y plane. Thefour white rectangles are the trap electrodes. Thetwo electrodes at x=0 are the AC electrodes,while the other two are the endcaps. The electricfield is at a minimum along the AC electrodes,which represent the trapping area. The field gen-erated by the endcaps will prevent the particlesfrom escaping the trap in the x direction.
(c) E2 along x, y and z axes. The red line represents E2 along they axis, the green line represents E2 along the z axis and the blueline represents E2 along the x axis. The maximum E2 is about3.51×1010 V2/m2 on the y axis, 1.4×1010 V2/m2 on the z axisand 1×1010 V2/m2 on the x axis.
Figure 4.6: Calculated electric field for the PCB trap. The AC and endcaps electrodes areheld at 300 V. The separation between the AC electrodes is 1.6 mm and thedistance between endcaps in x direction is 5.8 mm.
To summarise, this trap fulfils the two main criteria mentioned in section 4.1. The
trap shows a depth of about 5.5 ev (T=13×104 K) which is deeper by 60.5% com-
pared to the linear quadrupole trap, while the numerical aperture of this trap is
higher by 183% compared to the linear quadrupole trap. This is a significant im-
provement for fluorescence detection, apart from one direction of optical access
54
along the z-axis. This could be overcome by using a more complicated fabrication
process.
4.4 Stylus trapsThe stylus trap consists of two conical electrodes in a co-axial arrangement. It
has the ultimate optical access with a N.A. = 1, and has also been used to trap
micron-sized graphene particles [76, 77]. An AC voltage is applied to the outer
electrode, while the inner one is held at zero (DC) volts, as shown in Fig 4.7 and
Fig 4.8, a.
Figure 4.7: A 3D model of the stylus trap with conical electrodes in reference [76]. Thistrap consists of two conical electrodes in a coaxial configuration. The innerelectrode is DC electrode while the outer one is the AC. The apex of the innerelectrode is 200 µm higher than the outer electrodes, and the apex angle is 60.The outer electrode is a tapered cone with an upper diameter of 1.6 mm.
In order to examine the suitability of this trap, we have modelled the electric field
distribution using COMSOL Multiphysics. The distribution obtained is shown in
Fig 4.8, a. To determine the depth of the trap when the AC voltage amplitude is
held at 300 V, the trap electrode geometry was obtained from Kane et al. [76] The
E2 distribution shows an axial maximum of about 1.2× 108 V2/m2 and a radial
maximum of about 3.26×108 V2/m2 as shown in Fig 4.8, b and Fig 4.8, c. This is
99.6 % smaller than the linear quadrupole trap with endcaps, due to the non-linear
potential in the axial direction.
For matter-wave experiment, this trap shows a trap depth of about 0.07 eV
(T=0.15×104 K), which is lower by a factor of about 33 than the linear quadrupole
trap.
55
(a) Electric field contours. The dark blue triangle in the mid-dle represents the DC electrode. The right and left trianglesrepresent the cross section of the AC electrode.
2 0 2 1 2 2 2 3 2 4 2 5 2 6
1
3
0
2
4
E2 x108 (V
/m)2
A x i a l c o o r d i n a t e ( m m )(b) E2 in the axial direction. The maximum isabout 1.2×108 V2/m2.
R a d i a l c o o r d i n a t e ( m m )(c) E2 in the radial direction. The maximum isabout 3.26×108 V2/m2.
Figure 4.8: The electric field for the stylus trap with conical electrodes. The AC electrodesare held at 300 V and the DC at 0 V.
A second design, with similarly high N.A., was also studied to validate the low trap
depth of stylus traps [78, 79]. This design used cylindrical rather than the coni-
cal electrodes, and was surrounded by four cylindrical DC electrodes in a circular
arrangement for radial adjustment and confinement as shown in Fig 4.9.
56
Figure 4.9: Trap geometry of the styles trap with cylindrical electrodes as described byMaiwald et al. [78]. This trap consists of two cylindrical electrodes in a coaxialconfiguration, surrounded by four DC electrodes. The inner electrode is theground (GND) electrode, and the outer one is AC. The AC electrode has aninner radius of 267.5 µm, a thickness of 87.5 µm and a length of 1110 µm.The GND electrode has an inner radius of 50 µm and a thickness of 75 µm. Itis 500 µm higher than the AC electrode. The DC electrodes have a radius of75 µm, and are used to fine-tune the trapping potential generated by the ACand GND electrodes.
Using the same procedure as in the first design, the electric field contours and plots
were determined, and are shown in Fig 4.10. The maximum E2 for this design
is about 3.9× 107 V2/m2 in the axial direction, which is 67.5 % lower than the
previous design. We can conclude that the stylus trap has low trap depth compared
to PCB and linear quadrupole traps. Although it is possible to increase the trap
depth by raising the AC potential, this will increase the possibility of an electric
discharge between the trap electrodes, especially in vacuum, and of reaching the
breakdown voltage described by Paschen’s Law. We will discuss this further in
chapter 5.
57
(a) Electric field contours. The dark blue columns representthe plane cross section of the trap electrodes. The two centralcolumns represent the GND electrode, and adjacent two are theAC electrode which is shorter by 500 µm. The outer columns arethe DC electrodes.
1 . 8 2 . 3 2 . 82 . 0 2 . 5 3 . 00
2
4
6
8
1 0
E2 x107 (V
/m)2
A x i a l c o o r d i n a t e s ( m m )(b) E2 in the axial direction. The maximum isabout 3.9×107 V2/m2
- 1 . 0 - 0 . 5 0 . 0 0 . 5 1 . 00
1
2
3
4
E2 x108 (V
/m)2
R a d i a l c o o r d i n a t e ( m m )(c) E2 in the radial direction. The maximum isabout 3.8×108 V2/m2
Figure 4.10: Electric field for the stylus trap with cylindrical electrodes. The AC electrodesare held at 300 V and the GND and DC electrodes at 0 V.
In summary, the stylus trap satisfies only one of the two main criteria discussed in
section 4.1. Although it has a large N.A.(=1), due to a large opening angle, it has a
lower trap depth. As this is an issue that can be overcome by using more conven-
tional designs, we will not consider this trap further for use in these experiments.
58
4.5 ConclusionA range of electrical trap geometries have been modelled to determine their suitabil-
ity for diamond levitation, and for use in the matter-wave interferometry experiment
as described in chapter 1. These geometries included a linear quadrupole trap with
endcaps, a quadrupole trap constructed from a dual-sided circuit board (PCB), and
two stylus trap designs. The designs were chosen and evaluated on the basis of two
main criteria: (a) sufficient trap depth to hold particles for long periods, and (b)
high numerical aperture between trap electrodes for efficient NV fluorescence de-
tection. It was found that while stylus traps have optimum optical access (N.A. = 1),
the depth of this design is about 0.07 ev (T=1500 K) for a charge-to-mass ratio of
0.1 C/kg and a drive frequency of 1 kHz. The linear quadrupole trap and the PCB
trap both have much larger well depths with the same parameters, and the linear
quadrupole trap was approximately 33 times deeper than the stylus trap. The PCB
trap was found to be 2.5 times deeper than the linear quadrupole trap and both traps
had a high enough N.A. for fluorescence detection. The linear quadrupole trap had
an N.A. of 0.24 and the PCB trap had an N.A. of 0.68.
As the stylus trap is significantly shallower than the other traps, it will not be used
for trapping NV diamond. The construction and evaluation of the linear quadrupole
and PCB traps are described in chapter 5.
Chapter 5
Fabrication, loading and
characterisation of the Paul trap
5.1 OverviewIn this chapter we describe the construction and characterisation of the Paul traps,
PCB trap and linear Paul trap outlined in chapter 4. We also describe the particle
loading approaches used in the experiment, and an evaluation of the two traps that
were used.
5.2 Fabrication of implemented designsThe PCB trap and the linear quadrupole trap with endcaps were made to evaluate
their suitability for trapping. These traps were chosen for their deep trap depth and
the fact that they have sufficient N.A. for NV centre fluorescence detection.
5.2.1 PCB trap
The quadrupole trap was printed on a dual-sided PCB. Each side had a 70 µm
copper layer, coated with photoresist film. The fabrication was performed by using
a chemical etching process. The design layout was printed on transparent paper by
using a laser printer (see Fig 5.1), then placed onto the copper layer. Following this,
an ultra-violet lamp was used to transfer the design pattern onto the board using a
photoresist method, creating a mask to protect the desirable conductor. The board
was placed into an acid solution to etch and dissolve excess copper.
60
Figure 5.1: The design layout of the PCB trap printed on transparent paper.
In the region between the electrodes, where the particles should be trapped, a slot
was milled to allow particle loading and trapping.
5.2.2 Linear quadrupole trap with endcaps
The cylindrical rods for this trap are made of stainless steel, where each rod consists
of three main parts as shown in Fig 5.2.
DC electrode Nylon spacerAC electrode
Figure 5.2: A cross-section of one quadrupole rod. This consists of three main parts: anAC electrode, a DC electrode and spacers to isolate the two.
The AC electrodes are cylindrical rods with a length of 61 mm and a diameter of
4 mm. The rods are machined to reduce the diameter to 2 mm leaving the diameter
and length needed for the actual AC electrode in the centre as shown in Fig 5.3, a.
Two tubes of length 10 mm long, inner diameter of 4 mm and outer diameter of
5 mm were placed at each end of the AC electrode to form the DC endcaps
(Fig 5.3, b). The AC and DC electrodes were isolated from each other by two
nylon tube spacers inserted between them. This separated them by 2 mm as shown
in Fig 5.3, c. The final configuration and a photograph of the trap are shown in
Fig 5.3, d and Fig 5.4.
61
(a) AC electrodes. Four AC electrodes are con-structed using 61 mm cylindrical rods with adiameter of 4 mm. Each rod is machined toreduce its diameter to 2 mm over part of itslength, forming an AC electrode with a lengthof 7 mm.
(b) DC electrodes. Four tubes are placed ateach end of the AC electrodes to form the end-caps, each of which has a length of 10 mm, aninner diameter of 4 mm and an outer diameterof 5 mm.
(c) Nylon spacers. Eight nylon spacers areinserted between each AC and DC electrodesto mount the DC eletrode and provide electri-cal insulation. The AC and DC electrodes areseperated by 2 mm.
(d) The complete assembly. The electrodes arecoloured grey, and the red represents the nylonspacers.
Figure 5.3: A 3D model of the linear quadrupole trap with endcaps. The trap consist ofthree main parts: AC electrodes, DC electrodes and spacers.
62
Figure 5.4: A photo of the linear quadrupole trap with endcaps. The electrodes of this trapare made of stainless steel and separated by nylon spacers. Nylon holders wereused to hold the trap structure.
5.3 Loading methodsIt is important for this work that nanoparticles and microparticles are reliably loaded
into the traps. Some dissipation is required to reduce the kinetic energy of the par-
ticles within the trapping region, so that they have an energy lower than the trap
depth. Collisions with gas molecules are used to provide damping as the particles
are launched into the trapping region. In our experiments we have used three dif-
ferent methods to load the particles. These include the use of a piezo-speaker, an
nebuliser and an electrospray, where the particles are initially contained within a
liquid.
5.3.1 Piezo-speaker
A piezo-speaker from Murata Electronics (VSB35EWH0701B) was used to load
particles into the system. This is a ceramic piezoelectric speaker mounted in a
metal ring and driven by a function generator to create a vibrational motion. The
piezo-speaker was placed underneath the trap, located inside a vacuum chamber.
An applied voltage causes the piezo to vibrate at the drive frequency, which drives
63
particles toward the trap centre and overcomes any attractive force that the particles
have with the surface of the speaker.
This attractive force between the particles and the piezo disc is the van der Waals
(VDW) force, and it takes the form (assuming a spherical particle shape) [80–84]:
FVDW = 4πRγs, (5.1)
where γs is the effective solid surface energy and R is the particle radius.
The force generated by the piezo-speaker must be high enough to overcome the
VDW force and launch the particles. The minimum acceleration required to break
the VDW force between the piezo desk and a spherical particle of mass m is there-
fore [80–84]:
a =4πRγs
m=
4πRγs43πR3ρ
∝1
R2 , (5.2)
where m and ρ are the particle mass and density respectively. From this equa-
tion, we can see that the minimum acceleration required to break the VDW force
between the particles and the piezo disc surface increases as the size of the par-
ticles is decreases. For example, for a silica particle, where γs = 0.04 J/m2 and
ρ = 2650 kg/m3, we can calculate that the acceleration is 1.8× 1010 m/s2 for a
100 nm diameter particle, 1.8×108 m/s2 for a 1 µm diameter particle [83, 84], and
1.8×106 m/s2 for 10 µm diameter particle.
The acceleration is limited by the maximum force produced by the piezo-speaker,
which is equal to [81]:
Fpiezo = maap = maApω2p, (5.3)
where Ap and ωp are the piezo oscillation amplitude and its associated frequency
respectively, and ma is the effective mass. The maximum acceleration for the piezo-
speaker used in the experiment is approximately ap = Apω2p = 4.5×107 m/s2. This
64
value indicates the minimum size of silica particle that can be used, which is about
2 µm.
When larger particles of 10 microns are used, they are spread on top of the ceramic
piezo disc before being launched into the trap. For smaller sizes, the particles are
mixed with methanol, then sonicated for an hour in an ultrasonic bath, to avoid
clumping before the solution is decanted onto the piezo disk.
It was found that the particles were projected furthest when the signal generator
was set to sine wave at 20 V and 600 Hz. This is the resonant frequency of this
piezo. Using this piezo-speaker for loading generated a few issues. For example,
we experienced an electrical breakdown between the Paul trap electrodes during the
evacuation of the vacuum chamber. This breakdown occurred when the pressure
was in the tens of mbar range ( depending on the applied voltage between trap elec-
trodes, which was generally around 1 kV in our case). This breakdown is described
by Paschens Law, where the breakdown voltage VB between two electrodes in air is
given by:
VB =bPd
ln(aPd)− ln(ln(1+ γ−1)), (5.4)
where a = 1.125 mbar−1 mm−1, b = 27.375 V mbar−1 mm−1, γ = 0.02 are the
ideal case for air and stainless steel electrodes, P the pressure in mbar and d the
separation distance between the trap electrodes. The graph of equation 5.4 is known
as a Paschen curve and is shown in Fig 5.5. In our case, the Paschen curve for the
linear Paul trap and the PCB trap is shown in Fig 5.6. The minimum breakdown
voltage is about 260 V, which occurs at a pressure of about 3 mbar for the linear
Paul trap and about 6 mbar for the PCB trap.
65
100 101 102 103 104102
103
104
105
106
Pd (mbarmm)
VB(V)
Figure 5.5: Paschen curve for the ideal case in air. VB is the breakdown voltage betweentwo electrodes in air.
100 101 102 103102
103
104
105
106
107
Pressure (mBar)
VB(V
)
Figure 5.6: Paschen curve for the linear Paul trap and the PCB trap. The green line rep-resents the linear quadrupole trap with endcaps, where the separation distancebetween the trap electrodes d=3.5 mm, and the blue line represents the PCBtrap where d=1.6 mm. The minimum breakdown voltage VB is about 260 Vat a pressure of about 3 mbar for the linear quadrupole trap with endcaps andabout 6 mbar for the PCB trap.
The restricted breakdown voltage will be a barrier to applying the right trapping
voltages for high Q/m particles. For this reason, we looked for another approach
to overcoming the voltage breakdown issue, since we must trap at near-vacuum
66
conditions before evacuating to high vacuum. From the Paschen curve in Fig 5.5,
we notice that the breakdown voltage is high at atmospheric pressure and under
pressures of approximately 1 mbar. These two regions are adequate to operate the
Paul trap with a high trapping voltage.
5.3.2 Nebuliser
A nebuliser from Omron (MicroAir NE-U22) was used to load the Paul trap at at-
mospheric pressure, using the method of a passively vibrating mesh [85–87]. A
passively vibrating mesh nebuliser is an ultrasonic probe within the liquid that cre-
ates a strong vibration. This forces the solution containing the particles through the
multiple apertures of the mesh, each with a diameter of 3 µm.
The particles used in our experiment were delivered in a powder form or in 1 mg/ml
slurries in deionised water. Particles in powder form were mixed with methanol or
ethanol, then sonicated for an hour in an ultrasonic bath to avoid any aggregation of
the particles before decanting the solution into the nebuliser container. For particle
slurries in deionized (DI) water, droplets of slurry were mixed with methanol or
ethanol, and then sonicated in the same way as the powder solution.
It should be noted that the nebuliser performs inefficiently (i.e. at a low flow rate)
when particles are in acetone instead of methanol or ethanol. This could be a result
of clogging of the mesh holes due to fast evaporation of the solvent, causing subse-
quent particle aggregation.
Compared to the piezo-speaker described in the previous section, particles can be
trapped with a high trapping voltage at atmospheric pressure, before evacuating the
vacuum chamber to high vacuum. However, it is difficult to vary the trapping volt-
age under the breakdown voltage, while reducing pressure, without breaking the
stability condition for trapping.
5.3.3 Electrospray
A new method of loading was needed in order to prevent any clumping, and pro-
duce highly charged particles, while also being able to load the trap under vacuum
conditions. It was decided that an electrospray system was required for this as this,
67
although the piezo-speaker and the nebuliser were adequate for loading particles
(though not at Ultra High Vacuum (UHV) levels), the particles were not sufficiently
charged and occasionally clumped together. Electrospray works by forcing the liq-
uid that contains the particles through a hollow needle which is maintained at high
voltage. This high voltage produces a strong electric field that causes the solu-
tion droplets to break up (via strong Coulomb forces) and form a fine aerosol. The
droplet size is then further reduced to a steam of single highly charged nanoparticles.
A commercial electrospray deposition system (LV2 from MOLECULARSPRAY)
was used for loading particles under vacuum. This system as seen in Fig 5.7 con-
sisted of an electrospray emitter tip with a 100 µm diameter followed by a 0.25 mm
diameter capillary. A differential aperture was created using a 0.4 mm skimmer
cone and an exit flange. A PEEK microtee was then used to feed the solution from
a syringe tube to the emitter tip passing through a high voltage pin. The particle
solution was prepared using the same process as in nebuliser loading. This solution
was then decanted into the syringe barrel and delivered to the electrospray emitter
by gently pushing the syringe plunger. A high DC voltage with positive polarity
was applied directly to the solution through the microtee. The voltage was then
increased until a sustainable plume was created. This occurred at a voltage of ap-
proximately +3.5 kV, with about 2 mm separation between the emitter tip and the
entrance capillary.
The vacuum system has single stage pumping which used a scroll pump to reduce
pressure from atmospheric down to 1×10−1 mbar. The particles were guided to the
Paul trap by the differential pressure between this stage and the rest of the vacuum
chamber (which was about 1×10−3 mbar).
68
Paul Trap
+3.5 kV
Entrance capillary
Emittertip
Syringe
Gate valve
ScrollPump
P = 1X10-3
mbar
Skimmer
P = 1X10-1
mbar
Plume
Figure 5.7: A schematic diagram showing the electrospray setup and loading process. Anelectrospray deposition system (LV2 from MOLECULARSPRAY) was usedfor loading particles under vacuum. This system consisted of an electrosprayemitter tip with 100 µm inner diameter, an entrance capillary with a diameterof 0.25 mm and a 0.4 mm skimmer cone followed by an exit flange. This alsoacted as a differential pumping aperture. The solution delivery system con-sisted of a syringe with 1 mm diameter PEEK tubing and a PEEK microtee thatjoined the solution to the emitter via a high voltage pin. This high voltage pro-duces a strong electric field that disperses the solution to a fine aerosol, whichwas observed as a sustainable plume. The suction created by the differentialpressure pumping between the first pumping stage and the rest of the vacuumchamber guides the particles toward the Paul trap (from 1× 10−1 mbar at thefirst stage up to about 1× 10−3 mbar at the vacuum chamber). The inset is aphotograph of the electrospray plume.
69
5.4 Characterisation of Paul traps
5.4.1 PCB trap
5.4.1.1 Experimental set-up
The experimental scheme with a photograph of the final set-up is shown in Fig 5.8.
The particles are launched by the piezo-speaker; then pass through a skimmer,
which is a copper-less circuit board with a 1 mm hole in its centre, before finally be-
ing captured by the trap. The trapped particle is illuminated using a diode pumped
solid state laser (Verdi V10 from COHERENT) with a maximum output power of
10 W at 532 nm. The laser beam was coupled to an optical fibre using two mirrors
and a collimator, and the other end of the optical fibre was embedded within the
PCB through a 0.3 mm hole drilled from the board side toward the trap centre for
illumination. The light scattered from the particles was observed by a CCD camera
or a photomultiplier tube (PMT). The PMT was used to measure the fluctuations in
scattered light caused by the particle moving in the trap. This enabled measurement
of the trap secular frequency.
Figure 5.8: The experimental setup of the PCB trap. The trap was housed in a vacuumchamber, and a piezo-speaker used to launch the particles flowing through theskimmer toward the trapping region. The trapped particle was then illumi-nated by an optical fibre which was embedded within the PCB and coupled toa 532 nm laser using a collimator and two mirrors. For imaging, a CCD cam-era is connected to a zoom lens to observe the scattered light from the trappedparticle. The CCD camera could be replaced by a PMT to measure the trapsecular frequency. The inset is a photograph of the PCB trap inside the vacuumchamber.
70
5.4.1.2 Trap evaluation
We initially investigated the ability to trap different masses in air and under vacuum.
The first attempt used 10 µm diameter silica particles, but these were difficult to
load into the trap in air, so it was decided to use near vacuum conditions. After
several attempts under vacuum, we obtained successful trapping at 1.6×10−1 mbar.
All relevant operating parameters for this trap are listed in table 5.1. A photograph
of the trapped particle is also shown in Fig 5.9. Assuming that we were trapping 10
µm diameter silica, the charge is limited by the stability parameters and expected to
be between 1×1.6×10−19 C ≤ Q ≤ 24070×1.6×10−19 C to satisfy the stability
condition (qu < 0.908). For Q = 1× 1.6× 10−19 C the Mathieu parameters are
ar = 3.5×10−7 and qr = 3.7×10−5.
Table 5.1: Operating parameters for trapping 10 µm silica under vacuum in a PCB trap.
Figure 5.9: A photograph of 10 µm silica trapped under vacuum in the PCB trap. Thecharge is between 1×1.6×10−19 C≤Q≤ 24070×1.6×10−19 C based on theoperating parameters listed in table 5.1.
71
5.4.1.3 Limitation
The diverging light from the fibre tip dominated the scattered light from the trapped
particle. This made it difficult to measure the trap frequency. It was therefore
decided that a collimated beam should be used to illuminate the particle through
the top of the vacuum chamber. However, a major limitation of this trap was that
electrical breakdown occurred between the electrodes across the PCB substrate,
even after modifying the trap by adding an extra slot between AC and DC electrodes
as shown in Fig 5.10.
Figure 5.10: An extra slot was milled between the AC and DC electrodes in PCB trap.
It was decided to discontinue with this trap and, instead, continue with the linear
quadrupole with endcaps design, so that we could reliably trap particles without
breakdown. Although not used here, the PCB trap is currently being used for work
that relates to the cooling of silica nanoparticles in a hybrid system containing a
Paul trap and an optical trap [75].
5.4.2 Linear quadrupole trap with endcaps
5.4.2.1 Experimental set-up
The set-up for this trap is shown in Fig 5.12. The particles are loaded using three
different approaches. At the first stage, we used the piezo-speaker as we did for
the PCB trap, except that the piezo-speaker had a housing to guide particles toward
the trap centre as shown in Fig 5.11. At a later stage, an electrospray was used to
load the trap under vacuum and a nebuliser for loading at atmospheric pressure. For
illumination and detection, a similar set-up to that used in the PCB trap was used,
72
except that the optical fibre was attached to an adjustable collimator to focus the
laser beam used to illuminate the trapped particle.
Figure 5.11: A 3D model of part of the piezo housing.
Figure 5.12: The experimental setup of the linear quadrupole trap with endcaps. The trapwas housed in a vacuum chamber, and loaded using a piezo-speaker or anelectrospray. The piezo-speaker was placed underneath the trap (located in-side the vacuum chamber), and used to load the trap at atmospheric pressureor in near vacuum conditions, but not at high vacuum. An electrospray placedabove the trap was used for high vacuum loading. The trapped particle wasilluminated by a 532nm laser coupled to an optical fibre with an adjustablefocus collimator. The scattered light from the trapped particle was detectedby a CCD camera or a PMT.
73
5.4.2.2 Trap evaluation
Following the same process as with the PCB trap, we first tested the linear
quadrupole trap with endcaps at atmospheric pressure with two different sizes of
particle (10 µm and 2.5 µm) using the piezo-speaker. The only difference was that
the smaller particles were loaded using an ultrasonic nebuliser. The trap showed the
ability to levitate both sizes easily without an electrical breakdown between the trap
electrodes. The operating parameters are given in table 5.2. Based on these oper-
ating values, at atmospheric pressure, the charge for the 10 µm silica was between
1×1.6×10−19 C≤ Q≤ 63171×1.6×10−19 C, and for the 2.5 µm silica between
1×1.6×10−19 C≤ Q≤ 1542×1.6×10−19 C.
Under vacuum, the charge for the 10 µm silica was expected to be between
1× 1.6× 10−19 C ≤ Q ≤ 87504× 1.6× 10−19 C. An image of the trapped par-
ticle is shown in Fig 5.13. Note that the operating parameters for the 10 µm silica
varies between atmospheric pressure and under vacuum. This is due to the dif-
ference in the charge-to-mass ratio of the trapped particles, which are within the
stability region of the trap.
Figure 5.13: A photograph of a 10 µm silica particle trapped at 1.8×10−2 mbar in the lin-ear quadrupole trap with endcaps. The charge is between 1×1.6×10−19 C≤Q ≤ 87504 × 1.6 × 10−19 C based on the operating parameters given intable 5.2.
74
Table 5.2: Operating parameters for trapping 10 µm and 2.5 µm silica particles in the linearquadrupole trap.
Figure 5.14: A photograph of a 100 nm diamond trapped at atmospheric pressure in thelinear quadrupole trap with endcaps. The operating parameters are shown intable 5.3. The expected charge is between one or two elementary charges.
The next step was to trap 100 nm diamonds under vacuum to measure the secular
frequency. The piezo-speaker was replaced by an electrospray as discussed in sec-
tion 5.3 . We succeeded in trapping 100 nm diamond at 4× 10−3 mbar using the
trapping parameters given in table 5.4. It was now possible to measure the trap fre-
quency of the trapped particle by using the PMT. This was used to measure the fluc-
tuations in scattered light caused by the particle’s motion in the trap. The output cur-
rent of the PMT was connected to an oscilloscope, where it was converted to a time
varying voltage signal, and then recorded. By Fourier transforming the PMT signal,
it was possible to determine the trap frequencies as shown in Fig 5.15. We can see
that the driving frequency appears as the largest peak at 6.87 kHz with two side-
band peaks for ωz and ωx,y secular frequencies shifted by about 60 Hz and 209 Hz
respectively. A third peak, at 153 Hz, is related to the difference between the two
secular frequencies. We can estimate the particle radius R from the linewidth γg of
the secular frequencies peaks. This is given by [88, 89]:
γg =Γ0
2π=
12π
mGPrν
ρRkBT
(1+
π
8
). (5.5)
Here, Γ0 is the viscous damping factor caused by air, mG is the mass of the gas
molecule in amu unit, Pr is the pressure inside the chamber in Pascal, ρ and R are
the density and radius of the particle respectively, kB is the Boltzmann constant and
76
T is the temperature of the gas. The mean thermal velocity of the gas is given by
ν =√
8kBTπmG
. To define the linewidth γg of the secular frequencies peaks, we can fit
the power spectrum density (PSD) in Fig 5.15, where the time trace is in V2/Hz,
using the equation [90–92]:
S(ω) =2kBT
mγ2Γ0(
ω20 −ω2
)2+Γ2
0ω2, (5.6)
where m is the particle mass, γ is a calibration factor in V/m, ω0 is the secular
frequency and ω is the observed frequency. We used the following equation for
fitting:
S(ω) =A
(B2−ω2)2+C2ω2
+D, (5.7)
where A = 2kBT γ2Γ0m , B = ω0 and C = Γ0 are the fitting parameters and D is the
spectrum noise floor. The linewidth γg is about 3.21± 0.14 Hz. It is now possible
to calculate the radius of the trapped particle by using equation 5.5. This is about
R = 45±15 nm, this deviation in particle radius is due to uncertainty in the pressure
around the particle. We assumed that the pressure varied by a factor of two in the
range where the pressure transducer was located.
Table 5.4: Operating parameters for trapping 100 nm diamond at 4× 10−3 mbar in thelinear quadrupole trap.
Figure 5.15: The measured trap frequency after trapping 100 nm diamond under4× 10−3 mbar in the linear quadrupole trap. The red line shows the exper-imental PSD, and the blue line represents the fitted PSD using equation 5.7.The AC frequency appears as the largest peak at 6.87 kHz, with two side-band peaks for ωz and ωx,y secular frequencies, shifted by about 60 Hz and209 Hz respectively. The peak at 153 Hz represents the difference betweenthe ωx,y and ωz secular frequencies. The average linewidth γg of the secularfrequencies peaks is 3.21±0.14 Hz, where the spectrum noise floor is about3.56×10−10 V2/Hz.
We can estimate the charge, Q, of the trapped particle by using equations 3.27. The
charge in this case was Q = 2× 1.6× 10−19 C. Sine we have now calculated the
charge and mass of the trapped particle, we can calculate the Mathieu parameters,
which are ar = 1.4×10−4, qr = 8.76×10−2, az = 2.8×10−4 and qz = 4.9×10−3.
It should be noted that the maximum charge allowed, based on the operating param-
eters, is about Q = 30×1.6×10−19 C.
To validate the experimental results, we simulated the particle motion inside the
trap by using the equations of motion described in the chapter 3, after considering
geometrical efficiency factors. The secular frequencies were then calculated by us-
ing the power spectrum of the simulated particle trajectory. The power spectrum in
Fig 5.16 resembles that of the experiment, with ωx,y and ωz secular frequencies of
60 Hz and 209 Hz respectively.
78
0 50 100 150 200 2500.000
0.005
0.010
0.015
0.020
0.025
Frequency (Hz)
Amplitude
(a.u.)
Figure 5.16: Simulated secular frequency in the linear quadrupole trap. The first peakat 60 Hz is the ωz secular frequency. The second peak, at 209 Hz, is ωx,y.The parameters used for the simulation are Ω = 6.87 KHz, Vpeak−peak = 2 kV,U = +31 V, Q = 2×1.6×10−19 C and R = 45 nm.
5.5 ConclusionIn this chapter, we described the construction and characterisation of the linear
quadrupole trap and the PCB trap. We explored how these traps can be loaded
with particles using three different approaches: including a piezo-speaker, a nebu-
liser and an electrospray. We found that each techniques was suitable for loading
but an electrospray needs to be used for vacuum, due to electrical breakdown.
As the PCB trap was found to discharge between the electrodes across the PCB
substrate, this trap will not be considered further for the proposed experiments.
The linear quadrupole trap, however, was capable of trapping various particle sizes
at atmospheric pressure, and under vacuum, without an electrical discharge. It
was therefore selected as the preferred design for trapping NV diamond. We also
demonstrated the trapping of 100 nm NV diamond in vacuum, which is the first
step toward the proposed matter-wave experiment. In the next chapter we will de-
scribe how a microwave antenna was integrated into this trap, and investigate NV
photoluminesence for an NV diamond within this trap.
Chapter 6
Microwave excitation of NV diamond
6.1 OverviewAn important requirement for spin manipulation in nanodiamonds, and eventually
for matter-wave interferometry, is a strong microwave field within the vicinity of the
NV centre. For nanodiamonds on surfaces, this is usually accomplished by using a
small wire to act as a near field antenna in very close proximity (less than 100 mi-
crons) to the nanodiamond. Such proximity is problematic in a Paul trap, where the
distance between the trapped particle and the electrode structure is usually greater
than a millimetre. In addition, the other electrode of the Paul trap can interfere with
the propagating microwave fields within the trap.
In this chapter, we study the fields created inside the Paul trap using different an-
tenna structures. We begin by reviewing microwave antennas that have previously
been used for NV excitation. We then explore the use of a microwave antenna that
is part of the electrode structure of a linear Paul trap. The effect of the other Paul
trap electrodes on the distribution of magnetic field strength and power density is
then calculated.
6.2 Different design geometriesWhile several antenna geometries are available, our focus is on those which have
been used for NV excitation of a levitated diamond. A microwave antenna with
a loop shape has been implemented in optically levitated diamond experiments
[93, 94]. For example, Horowitz et al and Hoang et al. [93, 94] used a loop antenna
80
printed on a glass coverslip with an outer diameter of about 2 mm. In this study by
Hoang et al.,the antenna was placed 0.5 mm from the levitated nanodiamod, while
a study by Neukirch et al. [29] used a simple wire with a thickness of 25 µm. This,
according to the authers, was as close as possible to the levitated nanodiamond.
For NV diamonds levitated in ion traps, Delord et al. [42] used a wire antenna placed
150 µm from the levitated microdiamond in a needle trap. The wire’s position could
be adjusted to optimise the microwave power on the NV’s in the microdiamond. In
a further experiment, Delord et al. replaced the needle trap with a loop electrode
trap, where the loop was also the microwave antenna [31]. This trap had an outer
diameter of 2 mm and an inner diameter of 1.4 mm [31].
In order to understand the utility of these antenna, we simulated the magnetic field
strength ~H (A/m) and the electric fields ~E (V/m) around the antennas. The power
density was calculated using the Poynting vector, ~S = ~E × ~H. The magnetic field
strength and the power density were modelled using CST (Computer Simulation
Technology) MICROWAVE STUDIO. This software discretizes Maxwells integral
equation for a particular electrode geometry using the finite integration technique
(FIT) [95].
To determine the approximate MW magnetic field strength required for excitation
of the NV spin, we first need to know what Rabi frequency is required to drive the
spin. The Rabi frequency f , driven by a magnetic field, B0, takes the form [96]:
f =12
geµBB0/h, (6.1)
where µB is the Bohr magneton, ge is the free-electron g-factor and h is the Planck
constant. The minimum magnetic field that can be used is determined by setting
the Rabi frequency to 1/T2, where T2 is the spin decoherence time. T2 is of order
of microseconds for bulk diamond and hundreds of nanoseconds for nanodiamond.
The minimum magnetic field strength H0, can be then calculated using:
H0 = B0/µrµ0, (6.2)
81
where µr is the relative permeability of the material and µ0 is the vacuum per-
meability. We can now calculate the minimum magnetic field strength H0 to be
approximately 0.6 A/m for bulk diamond (assuming T2 = 100 µs) and 57 A/m for
nanodiamond (assuming T2 = 1 µs). Initially, a loop antenna of a similar design to
that used by Delord et al. [31] was modelled, as shown in Fig 6.1. This is also a
similar geometry to that used in the optical traps discussed above.
Figure 6.1: 3D geometry of the loop antenna. The outer radius is 1 mm and the inner radiusis 0.7 mm. This geometry is obtained from reference [31].
The absolute value of the magnetic field strength |~H|, and the power density for this
antenna, were calculated, and are shown in Fig 6.2. The input power was 0.5 W.
(a) Side view of H-field contours. (b) Top view of H-field contours.
Figure 6.2: Plots of the absolute value of the magnetic field strength contours created by aloop antenna with 0.5 W of input power. The colour signifies the absolute valueof the magnetic field strength in A/m. The antenna has an inner radius 0.7 mmand a thickness of 0.3 mm.
1 0 01 5 02 0 02 5 03 0 03 5 04 0 0 T r a p c e n t r e
H S
y ( m m )
H (A/m
)
2 . 0 x 1 0 5
7 . 0 x 1 0 5
1 . 2 x 1 0 6
1 . 7 x 1 0 6
2 . 2 x 1 0 6
2 . 7 x 1 0 6
3 . 2 x 1 0 6
y
S (W
/m2 )
Left p
art of
the lo
op an
tenna
Right
part
of the
loop
anten
na
z
x y
(b) |~H| and ~S along y axis.
Figure 6.3: Calculated absolute value of the magnetic field strength for the loop antennawith an outer diameter of 2 mm and an inner diameter of 1.4 mm. The inputpower is 0.5 W. The power density is about 559 kW/m2 at the trap centre and221 kW/m2 at a distance of 0.5 mm from the centre.
The power density in the middle of the loop is 559 kW/m2 (143 A/m) and
221 kW/m2 (88 A/m), at a distance of 0.5 mm away from the centre as seen in
Fig 6.3. This is the distance to the optically trapped particle reported by Hoang et
al. [94]. The magnetic field strength generated is 2.5 times greater than the mini-
mum strength required for nanodiamond excitation at the loop centre. It is 1.5 times
greater at a distance of 0.5 mm from the centre.
83
6.3 Microwave antenna integrated into the Paul trapFor simplicity, we studied the integration of an antenna within a linear Paul trap
and calculated the microwave field within the cylindrical electrodes of the Paul trap
described in section 4.2. This trap consists of four parallel cylindrical electrodes.
For trapping, an AC voltage is applied to the two opposing electrodes, and the other
two are grounded.
One of the grounded electrodes is now used as a microwave antenna (+ x axis). The
isolation of the grounded electrode from the high frequency microwave signal is
accomplished by use of an inductor. This antenna can be approximately described
as a monopole antenna, in terms of its geometry and the radiation pattern gener-
ated. The monopole antenna consists of a metal rod, mounted perpendicular to a
grounded surface. A MW source is connected between the lower end of the rod and
this surface. The oscillating current generates a microwave field with a resonant
frequency that depends on the length of the antenna, L. If this length is equal to a
quarter wavelength of the radiation L = λ/4 (for an infinite ground plane), then the
power radiated from the antenna is at a maximum. This is because the antenna is
impedance matched with the MW source.
Because, in our case, the antenna had no ground plane, and its length (which could
not be changed because of trapping requirements) was 28% of the ideal value. That
means we expect a power loss due to an impedance mismatch between the antenna
input and the source output. This meant that the antenna could not be used for effi-
cient far field propagation. However, as we are only interested in near field coupling,
the power transmission (coupling) between the antenna and the NV diamond might
be sufficient for NV excitation. We decided to evaluate this possibility. The com-
puted magnetic field strength for the electrodes, where the input microwave power
is 16 W, is shown in Fig 6.4. The absolute value of the magnetic field strength,
and the power density along the x,y and z directions, are shown in Fig 6.5,a,b and c
respectively, where the origin is the centre of the trap.
84
(a) Side view of H-field contours.
(b) Top view of H-field contours.
Figure 6.4: The absolute value of magnetic field strength contours of the monopole an-tenna at an input power of 16 W. The colour signifies the absolute value of themagnetic field strength in A/m. The antenna length is 7 mm and its diameteris 4 mm.
At the trap centre, the magnetic field strength is about 4 A/m and the power density
is about 18.4 kW/m2. This power density is lower than for the loop antenna (which
was 559 kW/m2, as seen in the previous section).
The centre of the trap in this configuration is 1.75 mm away from the MW electrode
surface (at the same distance using the loop antenna with 0.5 W input power is
H = 11 A/m and S = 3.6 kW/m2). We can also see that the absolute value of
the magnetic field strength decreases along the z axis, since the antenna feedline
is attached to the upper end of the electrode at +3.5 mm (the electrode length is
7 mm). In order to produce the same magnetic field strength for this structure as the
loop antenna, we would need to use an input power of 556 W.
85
- 6 - 4 - 2 0 2 4 6 8 1 0 1 2 1 4012345678
zy H
S
x ( m m )
H (A/m
)
MW an
tenna
T r a p c e n t r e
x
01 x 1 0 42 x 1 0 43 x 1 0 44 x 1 0 45 x 1 0 46 x 1 0 47 x 1 0 48 x 1 0 4
0 . 02 . 0 x 1 0 34 . 0 x 1 0 36 . 0 x 1 0 38 . 0 x 1 0 31 . 0 x 1 0 41 . 2 x 1 0 41 . 4 x 1 0 41 . 6 x 1 0 41 . 8 x 1 0 42 . 0 x 1 0 42 . 2 x 1 0 4
S (W
/m2 )
(b) |~H| and ~S along y axis.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 001234567
H S
z ( m m )
H (A/m
)
01 x 1 0 4
2 x 1 0 4
3 x 1 0 4
4 x 1 0 4
5 x 1 0 4
6 x 1 0 4
S (W
/m2 )
E l e c t r o d e l e n g t h
zy
x
(c) |~H| and ~S along z axis.
Figure 6.5: The absolute value of magnetic field strength and power density of a monopoleantenna at 16 W input power along x,y and z axes. At the trap centre, theabsolute value of the magnetic field strength is about 4 A/m, and the powerdensity is about 18.4 kW/m2.
86
Two further simulations were carried out in order to determine the effect of the
other trapping electrodes, as these will scatter radiation. Firstly, we performed a
calculation using the monopole antenna and the grounded electrode, with a sepa-
ration of 3.5 mm along the x axis. Fig 6.6 shows the variation in absolute value
of the magnetic field strength contours. We can see that, for this simple case, the
magnetic field strength is increased. From Fig 6.7, we can see that the magnetic
field strength is 4.73 A/m, and the power density is 23.92 kW/m2 at the trap centre.
This is an increase of about 30.4 % in the power density, compared to the monopole
microwave antenna/electrode alone. This is a result of constructive interference at
the trap centre.
(a) Side view of H-field contours.
(b) Top view of H-field contours.
Figure 6.6: Absolute value of the magnetic field strength contours of a monopole antennaat 16 W input power, after placing a second grounded electrode at a distance of3.5 mm.
87
- 1 5 - 1 0 - 5 0 5 1 0 1 50123456789
H S
x ( m m )
H (A/m
)
MW an
tenna
T r a p c e n t r e
Electr
ode
zy
x
0 . 0 01 . 5 0 x 1 0 4
3 . 0 0 x 1 0 4
4 . 5 0 x 1 0 4
6 . 0 0 x 1 0 4
7 . 5 0 x 1 0 4
9 . 0 0 x 1 0 4
S (W
/m2 )
(a) |~H| and ~S along x axis.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 00
1
2
3
4
5 H S
y ( m m )
H (A/m
)
E l e c t r o d e d i a m e t e r
zy
x
0 . 05 . 0 x 1 0 3
1 . 0 x 1 0 4
1 . 5 x 1 0 4
2 . 0 x 1 0 4
2 . 5 x 1 0 4
3 . 0 x 1 0 4
S (W
/m2 )
(b) |~H| and ~S along y axis.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 00123456 H
S
z ( m m )
H (A/m
)
E l e c t r o d e l e n g t h
zy
x
01 x 1 0 42 x 1 0 43 x 1 0 44 x 1 0 45 x 1 0 46 x 1 0 47 x 1 0 48 x 1 0 4
S (W
/m2 )
(c) |~H| and ~S along z axis.
Figure 6.7: Calculated absolute value of magnetic field strength and power density of amonopole antenna, after placing the second grounded electrode on x,y and zaxes. At the trap centre, the magnetic field strength is around 4.73 A/m withpower density around 23.91 kW/m2.
88
Next, we calculated the effect of all the trap electrodes. In this case, we con-
sidered the electrodes as ungrounded metal. The absolute value of the magnetic
field strength that is shown in Fig 6.8. Interestingly, the power density dropped to
about 13.7 kW/m2 and the magnetic field strength was around 3.4 A/m as shown in
Fig 6.9. This drop is about 25.5 % of the power density calculated for the single
monopole antenna. This configuration would require an input power of 654 W to
reach the absolute value of the magnetic field strength achieved in the experiments
of DeLord et al. [31].
(a) Side view of H-field contours.
(b) Top view of H-field contours.
Figure 6.8: Absolute value of magnetic field strength contours of a monopole antenna at16 W input power, when all quadrupole electrodes are taken into account.
89
- 1 5 - 1 0 - 5 0 5 1 0 1 50123456789 H
S
x ( m m )
H (A/m
)
T r a p c e n t r e
MW an
tenna
Electr
ode
zy
x
0 . 0 01 . 5 0 x 1 0 43 . 0 0 x 1 0 44 . 5 0 x 1 0 46 . 0 0 x 1 0 47 . 5 0 x 1 0 49 . 0 0 x 1 0 41 . 0 5 x 1 0 5
0 . 02 . 0 x 1 0 34 . 0 x 1 0 36 . 0 x 1 0 38 . 0 x 1 0 31 . 0 x 1 0 41 . 2 x 1 0 41 . 4 x 1 0 4
S (W
/m2 )
Electr
ode
Electr
ode
(b) |~H| and ~S along y axis.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 00123456
H S
z ( m m )
H (A/m
)
01 x 1 0 42 x 1 0 4
3 x 1 0 4
4 x 1 0 4
5 x 1 0 46 x 1 0 4
7 x 1 0 4
S (W
/m2 )
E l e c t r o d e l e n g t h
zy
x
(c) |~H| and ~S along z axis.
Figure 6.9: Absolute value of magnetic field strength and power density of the monopoleantenna after placing all quadrupole electrodes along the x,y and z axes. Theabsolute value of the magnetic field strength is about 3.4 A/m with a powerdensity of 13.7 kW/m2 at the trap centre.
90
To understand this drop in power density, we simulated the use of only two parallel
AC electrodes in the y direction with the microwave antenna. First, we simulated
the use of only one the electrodes, and then both. This comparison is shown in
Figure 6.10: A comparison of power density along x-axis for all the quadrupole configura-tions. At the trap centre, the power density for the microwave antenna aloneis about 18.4 kW/m2, 14.7 kW/m2 with one AC electrode, 11.4 kW/m2 withtwo AC electrodes, 24 kW/m2 with the grounded electrode and 13.7 kW/m2
with all electrodes.
The effect of placing the AC electrodes leads to a drop in the power and mag-
netic field strength because the two AC electrodes reflect the field and prevent it
from propagating into the centre of the gap between the electrodes. By placing one
AC electrode (the upper or the lower electrode) the power density drops by 20%
(H=3.52 A/m, S=14.7 kW/m2) and by placing both electrodes the drop is about
38% (H=3.03 A/m, S=11.4 kW/m2). In general, the AC electrodes decrease the
power density in the gap where the particle would be trapped. The grounded elec-
trode, in parallel with the same axis of the antenna electrode, increases the power
at the trap centre. Placing all electrodes will result in a drop of 25.5% in power
density, compared to a single electrode acting as the microwave antenna.
As noted in the previous section, more power could be obtained, in general, by us-
91
ing a loop antenna. The principal reason for this is that the feedline is connected to
one end of the antenna, and the other end is grounded. The monopole antenna, on
the other hand, has no grounded end. With this in mind, the idea of connecting the
two grounded electrodes in a loop, which would be closer a the loop antenna, seems
feasible. We examined this idea, using a linear Paul trap to form a loop antenna.
The absolute value of the magnetic field strength for this arrangement is shown in
Fig 6.11. Further, we can see in Fig 6.12 that the magnetic field strength increases
to 83 A/m with a power density of about 911 kW/m2 at the trap centre. This is an
increase of 480% in power density compared to the monopole antenna.
(a) Side view of H-field contours.
(b) Top view of H-field contours.
Figure 6.11: Absolute value of the magnetic field strength of a loop antenna formed fromtwo grounded electrodes of the quadrupole trap without AC electrodes. Inputpower was 16 W. The straight section of the loop is 7 mm in length with3.5 mm separation.
1 . 0 0 E + 0 52 . 5 0 E + 0 54 . 0 0 E + 0 55 . 5 0 E + 0 57 . 0 0 E + 0 58 . 5 0 E + 0 51 . 0 0 E + 0 6
S (W
/m2 )
L o o p t h i c k n e s s
z
y x
(b) |~H| and ~S y axis.
- 1 2 - 1 0 - 8 - 6 - 4 - 2 0 2 4 602 04 06 08 0
1 0 01 2 01 4 01 6 01 8 02 0 0
H S
z ( m m )
H (A/m
)
S t r a i g h t p a r t o f t h e l o o p a n t e n n a
Upper
part
of the
loop
anten
na z
y x
0 . 02 . 0 x 1 0 64 . 0 x 1 0 66 . 0 x 1 0 68 . 0 x 1 0 61 . 0 x 1 0 71 . 2 x 1 0 71 . 4 x 1 0 7
S (W
/m2 )
(c) |~H| and ~S along z axis.
Figure 6.12: Absolute value of magnetic field strength and power density of the loop an-tenna formed from two grounded electrodes of the quadrupole trap, withoutthe AC electrodes on x,y and z axes. The absolute value of the magnetic fieldstrength is about 83 A/m, and power density is about 911 kW/m2 at the trapcentre.
93
By placing the other two AC electrodes to form the complete trap, as shown in
Fig 6.13, we see a drop of 30.5% in power density, compared with the case with no
AC electrodes. The absolute value of the magnetic field strength at the trap centre
was 64 A/m, and the power density was 633 kW/m2 as shown in Fig 6.14. This
magnetic field strength is lower by a factor of 2, compared to the loop antenna used
by DeLord et al. [31], but is still high enough for NV nanodiamond excitation.
z
y
x
Feeding
port
Figure 6.13: Absolute value of the magnetic field strength of the loop antenna formed fromtwo grounded electrodes of the quadrupole trap, with 16 W input power. Thestraight part of the loop is 7 mm in length, and seperation is 3.5 mm.
0 . 02 . 0 x 1 0 54 . 0 x 1 0 56 . 0 x 1 0 58 . 0 x 1 0 51 . 0 x 1 0 61 . 2 x 1 0 61 . 4 x 1 0 61 . 6 x 1 0 61 . 8 x 1 0 62 . 0 x 1 0 6
S (W
/m2 )
Right
part
of the
loop
anten
na
Left p
art of
the lo
op an
tenna
T r a p c e n t r e
(a) |~H| and ~S along x axis.
- 1 0 - 8 - 6 - 4 - 2 0 2 4 6 8 1 00
1 0
2 0
3 0
4 0
5 0
6 0
7 0 H S
y ( m m )
H (A/m
)
1 . 0 0 E + 0 5
2 . 5 0 E + 0 5
4 . 0 0 E + 0 5
5 . 5 0 E + 0 5
7 . 0 0 E + 0 5
S (W
/m2 )
Electr
ode
Electr
ode
L o o p t h i c k n e s s
(b) |~H| and ~S along y axis.
- 1 2 - 1 0 - 8 - 6 - 4 - 2 0 2 4 602 04 06 08 0
1 0 01 2 01 4 01 6 01 8 02 0 02 2 0
H S
z ( m m )
H (A/m
)
S t r a i g h t p a r t o f t h e l o o p a n t e n n a
Upper
part o
f the lo
op an
tenna
0 . 02 . 0 x 1 0 6
4 . 0 x 1 0 6
6 . 0 x 1 0 6
8 . 0 x 1 0 6
1 . 0 x 1 0 7
1 . 2 x 1 0 7
1 . 4 x 1 0 7
1 . 6 x 1 0 7
S (W
/m2 )
(c) |~H| and ~S along z axis.
Figure 6.14: Absolute value of magnetic field strength and power density of the loop an-tenna formed from two grounded electrodes of the quadrupole trap along thex, y and z axes. The absolute value of the magnetic field strength is about64 A/m with power density of 633 kW/m2 at the trap centre.
95
6.4 ConclusionWe studied a range of different microwave antenna designs which are integrated
into a Paul trap which will be used for NV excitation of levitated NV diamond.
The minimum magnetic field strength required for NV excitation is approximately
0.6 A/m for bulk diamond assuming T2 = 100 µs (common for bulk diamond), and
57 A/m for nanodiamond assuming T2 = 1 µs. Our calculations showed that the lin-
ear quadrupole design, which incorporates a single antenna, generates a magnetic
field strength of 3.4 A/m using 16 W input power, which is available for the experi-
ment. This will be suitable at least for NV microdiamond, and may be suitable for
very clean nanodiamond.
We also showed that a higher field strength can be obtained by connecting the two
ground electrodes which, by forming a loop antenna, would radiate more microwave
power (64 A/m). However, integrating this design into our trap would require re-
engineering the trap geometry to fit the DC endcaps, so this will not be considered
further at this stage. This decision could be reconsidered at a later stage, if the
power generated by the monopole antenna is not sufficient for NV excitation. We
will investigate this experimentally, as described in chapters 7 and 8.
Chapter 7
NV photoluminescence with
temperature, laser intensity and gas
pressure
7.1 OverviewThis chapter describes a study of the photoluminescence (PL) of micron-sized dia-
mond embedded with NV centres, and placed within the Paul trap but on a micro-
scope glass slide. This study was undertaken to understand and evaluate the NV
excitation and detection system before carrying out levitation experiments. We be-
gin with a description of the microwave source used in this experiment, then outline
the ODMR experiments performed with it. Measurements of the NV photolumi-
nescence (PL) as a function of pressure, temperature and laser intensity are also
presented.
7.2 Microwave sourceFor the ODMR experiments a, voltage-controlled oscillator VCO (ZX95-3050A+
from Mini-Circuits) was used to generate microwave (MW) signals from 2150 to
3050 MHz, with an output power of 5 mW. This signal was amplified to 16 W
by using a high power amplifier HPA (ZHL-16W-43+ from Mini-Circuits). One
of the ground electrodes in the linear quadrupole trap was used as a microwave
97
antenna without further fabrication or modification of the design. As shown in
Fig 7.1, the output signal from the HPA was fed into the upper-right ground elec-
trode of the linear quadrupole trap. This meant that the two ports (GND and MW)
were simultaneously connected to the upper right electrode. A toroidal inductor
(MCAP109020040K-101MU from multicomp) with a 100 µH inductance was con-
nected between the other end of the ground electrodes, allowing DC currants to pass
through the electrode while blocking the MW signal.
MW
AC
ACGND
GND
&
MW
High Voltage
AC
GND
Vacuum chamber
VCO HPA
Figure 7.1: A schematic diagram of the microwave source. The set-up consists of a volt-age controlled oscillator (VCO) to generate microwave (MW) signals. Thesewere amplified to 16 W using a high power amplifier (HPA). The output sig-nal from the HPA was then fed into the upper-right ground electrode of thelinear quadrupole trap. A toroidal inductor with 100 µH inductance was con-nected between the other ends of the ground electrodes. This allows DC currentthrough the lower left electrode but blocks the MW signal.
7.3 ODMR detection set-upWe have demonstrated a cost-effective system for ODMR detection. This system
consisted of a microwave source (VCO and HPA), power splitter PS (ZFRSC-42-S+
from Mini-Circuits), an avalanche photodiode APD (SPCM-AQRH-13 from Perkin
Elmer), a data acquisition DAQ (USB-6003 from national instruments) and a fre-
quency counter (TF960 from Thurlby Thandar Instruments). This was controlled
using a Labview program. A schematic of this system is shown in Fig 7.2.
98
HPAVCO PS
USB Remote control
TF930Frequency Counter
Frequency
Count
Labviewprogram
Optical fibrePaul Trap
DAQ
APD
Figure 7.2: A schematic diagram of the ODMR detection set-up. The system consistedof a microwave source (VCO and HPA), power splitter (PS) and avalanchephotodiode (APD). Also shown is the data acquisition card (DAQ), frequencycounter and Labview program used to acquire the data. The Labview programwas connected to the DAQ to tune the VCO voltage, and the output frequencyof the VCO was split into two signals using the PS. One signal was amplifiedby the HPA, then fed to the MW electrode of the trap; the other signal wasconnected to the frequency counter to measure the output frequency. The dropin fluorescence intensity, due to microwave excitation when in resonance withspin transition, was detected by the APD. The output TTL pulses from the APDwere counted by the frequency counter, which was recorded on the computerrunning Labview. The program plots the MW frequency as a function of photoncount during the experiment.
The Labview program controlled the DAQ, which was tuned the MW frequency. It
also readout the photon counts that came from the APD via the frequency counter.
The program also controlled the VCO input voltage through the DAQ. The in-
put voltage was scanned between 7 V and 9 V in 0.01 V steps, which tuned
the frequency between 2.765±0.1 GHz to 2.975±0.1 GHz with a resolution of
1.05 ± 0.05 MHz. The VCO frequency output was split into two signal compo-
nents. One component was directed to the HPA, and then fed to the MW elec-
trode of the trap; the other component of the signal was connected to the frequency
counter which was used to measure its frequency. To record the photon count from
the APD, the TTL output signal was connected to the frequency counter. It should
be noted that this setup had a minimum update time of 0.3 s for reading the photon
count and the frequency input (total of 0.6 s for both), which meant that the system
99
took approximately 2 minutes to scan 200 MHz in 1 MHz steps. This is sufficient
to record an NV spectrum.
7.4 Experimental set-upMicron sized diamond powder (from Columbus Nanoworks) with a high concentra-
tion of NV centres was used for all experiments in this chapter. This relatively high
concentration allowed us to obtain strong photoluminescence signals. A syringe tip
was used to place and position the particles on a microscope cover slide. A drop
of ethanol was then added, to enhance adhesion between the particles and the slide.
The glass slide was then placed vertically between the trap electrodes, as shown
in Fig 7.3. A photographic image of the deposited particles is shown in Fig 7.4,
where the size of the deposited particles is about (20± 8)µm. This is determined
by calibrating the image against the known separation between the two Paul trap
electrodes, visible as the darker regions in Fig 7.4.
Vacuum chamber
PumpVacuum
gauge
Optical fibre
Collimator Optical fibre to
APD or spectrometer
MW
Glass coverslip
NF LPFND
ALAL
CCD
camera
Figure 7.3: A schematic of the experiment. A microscopic glass slide was inserted ver-tically between the trap electrodes. The upper-right electrode was used formicrowave excitation. An optical fibre connected to a collimator was used fora 532 nm laser excitation. The optical detection set-up consisted of two 50 mmachromatic lenses (AL) to collect the fluorescence emission, a neutral densityfilter (ND) with optical density = 2, a notch filter centred at 532 nm (NF), alongpass filter (LP) with a cut-off wavelength of about 610 nm, and an opticalfibre attached to the APD or the spectrometer. A CCD camera was used forimaging of the particle.
100
Figure 7.4: Photograph of the deposited particles, the pixel density is 4 µm/pixel, and thediameter is (20± 8)µm. The sub-images are close-ups of the deposited parti-cles with the laser on and off.
7.5 Recorded NV spectraIt was essential to identify the NV spectrum using a spectrometer (Andor Shamrock
303i spectrograph, with 1200 lines/mm grating and 40 nm bandpass), to ensure the
optical alignment and to test the microwave system used for ODMR. The spectrom-
eter split the incident light into individual narrow bands of wavelength, and each
band interacted with an array of pixels across the EMCCD camera for a given expo-
sure time. The EMCCD signal was converted into a voltage, which was then readout
as a digitized unit called a “count” which is divided by the exposure time. Fig 7.5
shows the PL spectrum obtained using 32 mW of laser power, with an intensity of
10.2 kW/cm2, and a 1.5 s exposure time of the EMCCD camera. The spectrum
was acquired between 500 nm to 900 nm. The zero phonon line (ZPL) of the NV−
within the sample was observed at 640 nm (Fig 7.6) as described in section 2.1. A
microwave field was then switched on and tuned to 2.867 GHz to get the maximum
reduction in the PL count. This reduction was calculated by subtracting the count
from the total area of the spectrum before and after applying the microwave field
with no background subtraction. The reduction in the total count was about 11%.
The optical contrast was expected to be in the range of 30%, this is because the NV
101
has been shown to be ≤ 30% of the time in the dark NV0 state [33, 34].
Figure 7.5: Spectrum of NV centre using a power intensity of 10.2 kW/cm2 and 1.5 s expo-sure time. The red line represents the spectrum acquired without the microwavefield, and the blue line represents the spectrum with the microwave field tunedto 2.867 GHz. The total PL count decreased by 11% with the microwave fieldswitched on.
6 3 0 6 3 5 6 4 0 6 4 5 6 5 0
2 5 0
3 7 5
5 0 0
6 2 5
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )
Figure 7.6: A close-up of NV− zero phonon line (ZPL). The laser intensity is 10.2 kW/cm2
and the exposure time is 1.5 s. The red line represents the spectrum acquiredwithout the microwave field, and the blue line represents the spectrum with themicrowave field. The PL intensity decreased by switching on the microwavefield at 2.867 GHz. This was due to excitation between the ms = 0 and ms =±1levels.
102
7.6 Optically detected magnetic resonanceAs described in sections 2.1 and 2.2, the ms = ±1 levels of NV diamond are off-
set from the ground state ms = 0 by 2.87 GHz and this is known as the Zero Field
Splitting (ZFS). To manipulate the spin state, we performed an electron spin reso-
nance (ESR) measurement by tuning a microwave field across the ESR resonance
to determine the ZFS. This resonance was observed as a reduction of PL count due
to the transition from the ms = 0 state to the ms = ±1 state and as a dip in the
recorded ESR spectrum. This process is called Optically Detected Magnetic Reso-
nance (ODMR). The ODMR detection set-up described earlier in this chapter was
used to perform this measurement. The ODMR spectrum was acquired between
2.77 GHz and 2.96 GHz with 1 MHz resolution at a laser intensity of 0.47 kW/cm2
as shown in Fig 7.7. The line shape of the ODMR spectrum could be described by a
Gaussian profile [23, 31], a Lorentzian profile [37, 97, 98] or a Voigt profile (a con-
volution of about 98% Lorentzian and 2% Gaussian) [99]. In our case, the shape is
a better fit to a Lorentzian rather than a Gaussian profile [97], especially for power
broadened ODMR as in our case [100]. It is worth mentioning that splitting of a
few MHz could be detected with no external magnetic field, due to the local crystal
strain field [37]. Therefore, when fitting the data, the ODMR dip was considered as
the midpoint of the two split dips. The ZFS was found to be 2.872±0.001 GHz as
expected for NV diamond at room temperature [32]. The ODMR contrast was about
17 ± 1 % which is expected since NV spends up to ≤ 30% of the time in the dark
NV0 state [33]. The peak width was about 70 MHz, which is considerably wider
than expected from the literature; typically around 5 MHz [37]. This broadening
could be related to the high RF power generated by the microwave source [97,101],
which is not adjustable in the current set-up.
103
2 . 8 0 2 . 8 5 2 . 9 0 2 . 9 50 . 7 5
0 . 8 0
0 . 8 5
0 . 9 0
0 . 9 5
1 . 0 0
Norm
alize
d PL c
ount
F r e q u e n c y ( G H z )
Figure 7.7: An ODMR spectrum at 0.47 kW/cm2. The red dots represent the measureddata while the red line represents a Lorentzian fit to the data. The shaded areais the error envelope for 2σ which is about ± 1 %. The plot is normalisedby the maximum output photon count, which represents the number of TTLoutput pulses from the APD within 1.5 s of exposure time. The ODMR dip wasconsidered as the midpoint of the two dips when fitting the data.
We also obtained spectra under three different laser intensities, but with the same
microwave field as seen in Fig 7.8. It can be seen that the ODMR contrast reduces
with higher laser intensity with a slight shift of the ZFS frequency below 2.87 GHz
as seen from the single Lorentzian fit in Fig 7.8. This data is summarised in table
7.1. The ZFS frequency shift and the reduction of the ODMR contrast is caused by
heat generated from the laser. This heat will increase the NV diamond temperature
and consequently increase the probability of non-radiative transition through the
singlet state.
Table 7.1: ZFS frequency and ODMR contrast obtained from ODMR spectra in Fig 7.8.
Figure 7.8: The ODMR spectrum at laser excitation intensities of 0.47, 3 and 10.2 kW/cm2.Each line is a single Lorentzian fit of the data, and the shaded area is the errorenvelope for 2σ . A few MHz splitting was detected, with no magnetic field,due to internal strain. The ODMR dip was considered as the midpoint of thetwo dips when fitting the data. The ODMR contrast reduces as laser intensityincreases, and there is a slight shift in resonance frequency. This is probablydue to the heat generated in the crystals from absorption of the laser.
105
7.7 Zeeman splitting of ODMR spectraIn section 2.2, we discuss the effect of an external magnetic field (B) on the ODMR
spectra. In the presence of a magnetic field, the degeneracy of the ms = ±1 is lifted
by the Zeeman effect. The ODMR spectrum has two dips due to each ms = ±1
states. The separation in frequency (4ν) between these two dips changes with the
strength of the applied magnetic field B as:
4ν = 2γB, (7.1)
where γ = 28024.951 MHz/T is the electron gyromagnetic ratio.
The ODMR spectrum for different laser intensities, with an external magnetic field,
is shown in Fig 7.9. The magnetic field was generated by a neodymium magnet
placed inside the vacuum chamber. As in the field-free case above, a reduced
contrast occurs with increased laser intensity (7.6). The data showing the ESR fre-
quency shift in both peaks, resulting from increasing laser intensity, is summarised
in table 7.2.
Table 7.2: Frequency resonances and ODMR contrast obtained from split ODMR spectrain Fig 7.9.
Figure 7.9: Zeeman splitting of an ODMR spectrum at 0.47, 3 and 10.2 kW/cm2. The dotsrepresent the experimental data, the dashed lines represent the a Lorentzian fitof each peak, while the solid lines represent the sum of two Lorentzian peakscorrelated with the ms = +1 and ms = −1 states. The shaded area is the errorenvelope for 2σ . The ODMR spectrum experiences a Zeeman-shift due to anexternal magnetic field generated by a neodymium magnet. The two fluores-cence frequencies correlated with the ms = ±1 states appeared as two dips inthe spectrum. As with a typical ODMR spectrum, the heat generated from thelaser reduces the ODMR contrast with a slight shift in the resonance frequency.
107
A comparison of the ODMR spectrum with and without an external magnetic field
is shown in Fig 7.10. The separation in frequency between the two spin states is
48 MHz, which is consistent with that produced by a magnetic field of 0.84 mT
using equation 7.1. The magnetic field strength was measured using a commercial
Gaussmeter (Hirst Magnetics GM07), with the probe placed between the trap elec-
trodes, in the same position as with the diamond sample. The measured magnetic
field strength was 0.82±0.04 mT, which was consistent with the calculated value.
All measurements were carried out at this fixed magnetic field strength. We have
assumed that most of the NVs within the sample have the same orientation, and that
their axes are parallel to the magnetic field. If this were not the case, we should
observe more than two dips in the ODMR spectrum. This conclusion is consistent
with the observation reported by Fukui et al. [39]. It is probable, however, that other
orientations exist at lower density, and we cannot resolve them.
2 . 8 0 2 . 8 5 2 . 9 0 2 . 9 50 . 8 5
0 . 8 8
0 . 9 1
0 . 9 4
0 . 9 7
1 . 0 0
B = 0 TB » 0 . 8 4 m T
Norm
alize
d PL c
ount
F r e q u e n c y ( G H z )
Figure 7.10: Comparison between the ODMR signal with and without external magneticfield at 3 kW/cm2. The two resonance dips are separated by 48.3 MHz, whichis the same as the seperation in a magnetic field strength of 0.84 mT. Thecontrast dropped from 15 ± 0.5% with no magnetic field, to 7.4 ± 0.2% witha magnetic field.
108
7.8 Photoluminescence as a function of pressureThe PL at different laser intensities and pressures was studied. The average PL level
as a function of laser intensity (also called a PL saturation curve) is illustrated in
Fig 7.11. These measurements were taken with an APD and each data point is an
average of 10 measurements with a 1.5 s exposure time per measurement.
1 3 5 7 9 1 1 1 3 1 5 1 7 1 9
1 . 0
3 . 0
5 . 0
7 . 0
9 . 01 ´ 1 0 + 3 m b a r4 . 5 ´ 1 0 - 1 m b a r2 . 2 ´ 1 0 - 4 m b a rF i t u s i n g I P L ( L ) = I ( L / L + L s a t ) U n e x p e c t e d P L i n c r e a s e d u e t o
i m p u r i t i e s i n t h e s a m p l e P o l y n o m i a l f i t L i n e a r f i t
PL in
tensit
y (´1
06 coun
ts/s)
L a s e r i n t e n s i t y ( k W / c m 2 )Figure 7.11: PL saturation curves at different pressures. Lowering the pressure reduces the
PL count rate until it saturates. At this point, it starts quenching due to theincreased temperature of the diamond lattice. As the background gas acts asa heatsink, lower pressure causes an increase in internal temperature, whichincreases the probability of the non-radiative transition through the singletstate. The data at atmospheric pressure was fitted using equation 7.2, whereI∞ = 27 Mcounts/s and Lsat = 31 kW/cm2. This equation shows good agree-ment with the experimental data at atmospheric pressure, though not when thePL intensity starts to decline under vacuum. The data was fitted using a thirdorder polynomial, which was found to be in very good agreement with thethree different pressure levels. The unexpected increase due to impurities inthe sample was not included in the fitted data.
At atmospheric pressure, the fluorescence intensity increases linearly as a func-
tion of laser intensity. It then starts to saturate when the intensity is higher than
6.5 kW/cm2, due to an increased probability of non-radiative transitions through
the singlet state. This behaviour at atmospheric pressure has been previously ob-
109
served and the relationship between PL intensity IPL and laser intensity is given
by [102] :
IPL(L) = I∞(L
L+Lsat), (7.2)
where I∞ is the maximum intensity of the PL emission for infinitely high laser in-
tensity and Lsat is the laser intensity related to I∞.
It was found that, under vacuum conditions, this effect was enhanced. By reducing
the pressure to 4.5× 10−1 mbar, from atmospheric pressure (1000 mbar), the PL
intensity reaches a maximum at 11.5 kW/cm2, but then decreases at higher laser
intensities. This suppression of PL at higher intensities has been observed before,
and is believed to be due to the increase in temperature of NV diamond up to about
300 - 400 C (573.15 K - 673.15 K) [103, 104]. To verify that this reduction in
intensity is due to the NV, and not the excitation of impurities in the diamond or
optical components, we acquired spectra with a spectrometer, and measured how
the spectrum changed with laser intensity. This was carried out with two dif-
ferent laser intensities, each at two different pressures (atmospheric pressure and
4.5×10−1 mbar). These results are shown in Fig 7.12. The resulting spectra showed
similar behaviour to the PL saturation curves (Fig 7.11) at the same pressures.
The NV centre has two charge states NV− and NV0, and both can be transformed to
the other state [105, 106]. It is possible that laser excitation can prduce this change
in state. If so, we should observe a shift of the NV− spectrum (PL spectrum centred
at 680 nm) due to the increased presence of NV0 (PL spectrum centred at 610 nm)
states, so that the whole PL spectrum is centred between 610 nm and 680 nm in-
stead of 680 nm. However, by normalizing, and comparing two PL spectra at both
intensities (one at maximum intensity and the other at reduced intensity), we can see
(Fig 7.13) that there was, in fact, no shift in the spectrum. This is in agreement with
the findings of Plakhotnik and Chapman [103] who did not observe any transforma-
tion of the charge state with high laser intensity.
Figure 7.12: A PL spectrum at atmospheric pressure and at 4.5× 10−1 mbar. The spectrawere acquired for two laser intensities: 10 kW/cm2 and 15 kW/cm2, chosen torepresent the spectrum at the PL saturation level, and where the PL intensitywas quenched at 4.5×10−1 mbar. The intensities of these spectra demonstratethat the saturation and quenching of the fluorescence was connected to the NVdiamond.
6 0 0 7 0 0 8 0 0 9 0 00 . 0
0 . 2
0 . 4
0 . 6
0 . 8
1 . 0
Relat
ive PL
sign
al
W a v e l e n g t h ( n m )
Figure 7.13: A comparison of two PL spectrum profiles at 4.5× 10−1 mbar. The red linerepresents the spectrum at 10 kW/cm2 while the blue line represents the spec-trum at 15 kW/cm2. Each spectrum is normalized, so that the spectrum peakhas a high point of 1. The first spectrum is at 10 kW/cm2, which is the satu-ration level, and the other is at 15 kW/cm2 where the intensity declined. Noshifts of spectra was observed, because there is no change in the charge statefrom negative NV− to neutral NV0.
111
A further investigation was carried out at the higher vacuum level of at
2.2× 10−4 mbar. In this case, the saturation level was reached at 10 kW/cm2
followed by a reduction in the PL. However, at higher laser intensity, the fluores-
cence count increased, a result which has not been previously observed. A wide
scan of the PL spectrum indicates that this rise in the PL is not associated with
the characteristic NV spectrum (Fig 7.14), but is probably due to impurities in the
Figure 7.14: A PL spectrum at 2.2× 10−4 mbar. Increasing the laser intensity reducesthe fluorescence count. At higher laser intensity, the spectrum showed anadditional emission band between 750 nm and 1050 nm. This increased withlaser intensity, and is probably related to impurities in the sample.
Reducing the pressure decreases the PL count rate until saturation point is reached.
A further increase in laser intensity quenches the fluorescence, probably due to in-
creased temperature of the diamond. The background gas acts as a heatsink, so that,
at lower pressures, the internal temperature increases, thus increasing the probabil-
ity of non-radiative transitions.
From Fig 7.11, we can see that equation 7.2 agrees with the PL saturation curve at
atmospheric pressure, but does not describe the PL intensity in vacuum, where it
(the PL intensity) starts to decline at high laser intensities. We found that the pho-
112
toluminescence intensity IPL could be described as a function of the laser intensity
L by the third degree polynomial:
IPL(L) =C0 +C1L+C2L2 +C3L3, (7.3)
where C0 = 0 is the intersect, set at 0 count/s (no PL at zero laser intensity), and
C1, C2 and C3 are the coefficients in the polynomial fit, listed in table 7.3. Note
that derivative of this cubic function shows how photoluminescence intensity varies
with laser intensity as shown in table 7.4.
Table 7.3: Coefficients in the polynomial fit for the saturation curves in Fig 7.11.
Figure 7.15: ODMR spectra at atmospheric pressure with varying laser intensity. The datawas fitted using a Lorentzian function. The ESR frequency decreases withincreasing laser intensity. The ODMR contrast, however, increased initially,but peaked up at 3 kW/cm2, then decreases at higher laser intensities. Allvalues are shown in table 7.5.
115
Table 7.5: ZFS frequency and ODMR contrast obtained from Fig 7.15.
Figure 7.16: ZFS frequency for various laser intensities, at atmospheric pressure. The fre-quencies were obtained from the ODMR spectrum in Fig 7.15. The shift ofZFS frequency was about -9 MHz over the entire range.
116
Figure 7.17: ODMR spectra at 4.5× 10−1 mbar for different laser intensities. The ZFSfrequency and contrast decreased with increasing laser intensity. By reducingthe pressure, the heat transfer rate between the sample and the surroundinggas was decreased, resulting in a large shift in the ZFS frequency. All valuesare shown in table 7.6.
117
Table 7.6: ZFS frequency and ODMR contrast obtained from Fig 7.17.
Figure 7.18: The ZFS frequency for various laser intensities at 4.5× 10−1 mbar. The fre-quencies were obtained from the ODMR spectrum in Fig 7.17. The shift inZFS frequency was about -42.2 MHz over the entire range.
5 k W / c m 22 . 7 9 2 . 8 2 2 . 8 5 2 . 8 8 2 . 9 1 2 . 9 4
F r e q u e n c y ( G H z )
8 k W / c m 2
Figure 7.19: ODMR spectra at 2.2×10−4 mbar for different laser intensities. The ZFS fre-quency and contrast decrease with increasing laser intensity. The temperatureincreased significantly due to the reduction in the surrounding gas, which actsas a heatsink. All values are shown in table 7.7.
Table 7.7: ZFS frequency and ODMR contrast data from Fig 7.19.
Figure 7.20: The ZFS frequency for various laser intensities at at 4.5× 10−1 mbar. Thefrequencies were obtained from the ODMR spectrum in Fig 7.19. The shift inZFS frequency was about -36.6 MHz over the entire range.
7.10 Temperature dependence of ZFSIn previous sections, we have demonstrated that a decrease in pressure quenches
NV photoluminesence and reduces the ODMR contrast with a shift of the ZFS fre-
quency. This is probably a result of the increased temperature of the NV diamond
when it is exposed to the 532 nm excitation laser. The temperature dependence of
ZFS has been previously studied. These models are grouped into three different
temperature regions, based on how the zero field splitting varies [49, 50, 52].
In the low temperature region from 5.6 K to 295 K, ZFS changes with temperature
are described by a fifth degree polynomial. The change in ZFS is about 7 MHz (from
2.877 GHz to 2.87 GHz) in this temperature range [50]. In the middle temperature
region, between 280 K and 330 K [49], ZFS varies linearly with temperature but
with a negative slope equal to dD/dT =-74.27 kHz/K, where D represents the ZFS
frequency. Third, in the high temperature region from 300 K to 700 K, the variation
of D with temperature is described by a cubic polynomial function with a thermal
120
shift that varies from dD/dT =-80 kHz/K at 300 K up to dD/dT =-170 kHz/K at
700 K. The shift in D between these two temperatures is about -56 MHz [52].
From these three regions we can see that D(T ) shifts up from 2.87 GHz when the
temperatures is below 300 K, and shifts below 2.87 GHz when the temperature is
higher than 300 K.
Our measurement in the previous section shows that the ZFS frequency shifts below
2.87 GHz when we increase laser intensity, at either atmospheric pressure or under
vacuum. This decrease in the ZFS frequency indicates that we are operating in the
high temperature region discussed above, and that the temperature is above 300 K.
The ZFS frequency parameter denoted as D in this region is described by a cubic
Figure 7.21: Temperature dependence of laser intensity at different pressure levels. Theblack dots represent the data at atmospheric pressure, the blue dots at4.5× 10−1 mbar and the green dots at 2.2× 10−4 mbar. The solid line rep-resents a linear fit of the data. All linear fits intersect the temperature axisaround 320±20 K.
As the frequency shift due to the local strain has been determined, equation 7.4 can
be used to calculate the temperature from the ZFS frequency as shown in Fig 7.22.
For a given laser intensity, lowering the pressure decreases the heat transfer rate be-
tween the NV diamond and the surrounding gas. From Fig 7.23, we can determine
the temperature at fixed laser intensities for three different pressure levels. For ex-
ample, at a laser intensity of 8 kW/cm2, the ZFS frequency shifts of about -35 MHz
from atmospheric pressure down to 2.2×10−4 mbar, which raises the temperature
Figure 7.22: Deduced temperature from ZFS at different pressure from a change in laserintensity. The red line represents the cubic polynomial fit of the data points.The temperature is derived from the ZFS frequency using equation 7.4. In-creasing the laser intensity from 0.5 kW/cm2 to 8 kW/cm2 leads to a ZFS fre-quency shift about -8.5 MHz at 1000 mbar, -14.4 MHz at 4.5×10−1 mbar and-37 MHz at 2.2×10−4 mbar. This raises the temperature by 92 K, 116 K and259 K respectively.
Figure 7.23: Comparison of the ZFS frequency and corresponding temperature from achange in laser intensity at different pressure levels. The black dots rep-resent the data at atmospheric pressure, the red dots represent the data at4.5× 10−1 mbar, the blue dots represent the data at 2.2× 10−4 mbar. Theerror bars show the uncertainty in frequency. The temperature axis is calcu-lated using equation 7.4 for each corresponding frequency. Lower the pres-sures probably decrease the heat transfer rate between the NV diamond andthe surrounded gas, which increases the temperature of the diamond lattice.The result is a large shift in ZFS frequency. Between atmospheric pressureand 2.2× 10−4 mbar, the temperature was increased by 119 K at 3 kW/cm2,191 K at 5 kW/cm2 and 247 K at 8 kW/cm2.
7.11 ODMR contrastAs discussed in section 7.9, the temperature change not only affects the reso-
nance frequency but also the ODMR contrast. Based on temperatures correspond-
ing to ZFS frequency, we can draw a relationship between the temperature and
ODMR contrast. It can be seen from Fig 7.24 that, by increasing laser intensity to
3 kW/cm2, the contrast increases from 9.8% to about 11.7%. The temperature at
3 kW/cm2 was about 366 K. At higher laser intensities, the contrast starts to de-
crease until it reachs 9.43% at 384 K, where the laser intensity was 10 kW/cm2. Un-
der vacuum, the contrast slightly increases between laser intensities of 0.5 kW/cm2
124
to 3 kW/cm2, peaking at 11.7%. It then falls with temperature down to 3− 5% as
seen in Fig 7.25 (at 4.5×10−1 mbar) and Fig 7.26 (at 2.2×10−4 mbar).
Figure 7.24: Effect of temperature change on ODMR contrast at atmospheric pressure. TheODMR contrast was 9.8% at room temperature, peaking at 11.7% at 366 K.It then decreased to 9.43% at 384 K.
Figure 7.26: The effect of temperature change on ODMR contrast at 2.2×10−4 mbar. TheODMR contrast fell from 8.4% at 366 K to 5% at 625 K.
In summary, from the observed measurements, the ODMR contrast reached its peak
at about 380±35 K by using 3 kW/cm2 of laser intensity at atmospheric pressure
and 4.5×10−1 mbar. It then fell, due to the PL saturation associated with increased
probability of non-radiative transition.
7.12 ConclusionIn this chapter we demonstrated optically detected magnetic resonance (ODMR) of
NV centres in diamond, placed in a linear quadrupole trap using a cost effective
detection set-up. We showed experimentally that using one of the linear trap elec-
trodes as a monopole antenna, as described in chapter 6, is feasible, and that the
power generated was sufficient for NV excitation. NV fluorescence from microdi-
amond was used to investigate how NV photoluminesence varies as a function of
laser intensity, temperature, and gas pressure. We found that, at high laser inten-
sities, the photoluminescence of NV− (the PL saturation curve) is quenched under
vacuum, due to the increased probability of non-radiative transition. We also found
that the reported theoretical expression for PL saturation, IPL(L) = I∞(L
L+Lsat), de-
scribes well the PL at atmospheric pressure but not in vacuum. A third degree
polynomial was shown to fit the PL saturation curve under any pressure level. The
126
quenching was a result of the increased temperature of the NV diamond, and can
be estimated by measuring the changes in resonance frequency. These measure-
ments showed that the NV diamond sample reached a temperature as high as 657 K
at 4.5× 10−1 mbar, with an ODMR contrast of 3.2%. The results also showed
that temperature increased by 247 K when lowering the pressure from atmospheric
pressure down to 2.2×10−4 mbar at a laser intensity of 8 kW/cm2. The results also
indicate that the contrast reduces when the temperature is higher than 380±35 K.
The findings reported in this chapter therefore provide the fundamentals of how
temperature changes due to laser intensity and pressure can drive the non-radiative
decay and influence the photoluminesence properties significantly. This will be an
important aspect when considering the use of NV diamond to carry out the macro-
scopic matter-wave experiments, as optical readout of the spin state is based on the
PL intensity of the NV centres, which is used to evidence the spatial superposition.
It is important for this experiment to have fast MW pulses to manipulate the spin
state. These pulses should be tuned to a given frequency that is equal to the ZFS
frequency. As we showed in this chapter, this frequency will shift with the tem-
perature changes induced by the optical field. This means that before starting the
experiment, we must measure the ZFS frequency. This frequency should not change
during the experiment.
The results in this chapter have helped us to understand the effect of heating on pho-
toluminescence properties, especially under vacuum, which will be the main issue
for levitated NV diamond as will described in the next chapter.
Chapter 8
Observing NV photoluminescence in
the Paul trap
8.1 OverviewIn previous chapters, we have described the construction and characterisation of the
Paul trap, as well as the optical and microwave systems used to perform ODMR on
an NV nanodiamond. In this chapter, we describe the levitation of diamond particles
within the trap and its prospects for future in-trap matter-wave experiments.
8.2 Sample characteristicsIt was decided to begin trapping experiments with micro-diamond rather than nan-
odiamond, as it is easier to observe fluorescence. Initial experiments utilised 1 µm
diamond in powder form (from Columbus Nanoworks). A mixture of these par-
ticles with ethanol was introduced into the trap by a nebuliser (Omron MicroAir
NE-U22). However, though particles were easily trapped, no PL was observed. Ini-
tially, it was thought that this could be due to the presence of other particulates in
the powder supplied to us, so the PL spectrum of the samples was studied, using a
RENISHAW inVia Raman Microscope with 514 nm excitation. The samples avail-
able for trapping were 1 µm and 100 nm diamonds from Columbus Nanoworks in
powder form, and 100 nm particles from Adamas Nanotechnologies delivered as
1 mg/ml slurries in de-ionized (DI) water. The first sample examined was 1 µm
diamond, which as mentioned above, did not produce a detectable PL signal. The
128
nebuliser used for introducing the particles into the trap was also used to deposit the
particles on a microscope glass slide. This was done to determine whether we could
observe any NV PL from the glass slide, which should contain many more than
the single particle levitated in the trap. Several clusters of the same 1 micron size
diamonds deposited on the microscope slide, and the PL spectra was collected by a
10X microscopic objective. The resulting spectrum showed a broadened peak be-
tween 630 to 850 nm as seen in Fig 8.1, which is within the range of the broadened
peak usually observed in NV diamond. However, there was no peak for NV− ZPL
at 640 nm. Other diamonds, from different parts of the sample, showed a similar
spectrum to the first. From these spectra, we concluded that this batch of microdia-
monds probably does not contain NV centres. The broadened peak observed seems
very similar to the fluorescence from an empty glass slide, as can be seen from the
in Fig 8.2. The glass slide was therefore replaced with a quartz one. After this, the
spectra showed no signs of the glass fluorescence or the NV spectrum as shown in
Fig 8.3.
5 5 0 6 0 0 6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 0
2 0
4 0
6 0
8 0
1 0 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.1: PL spectra of 1 µm in diameter microdiamond particles from Columbus
Nanoworks. The particles were mixed with ethanol and then deposited ontoa microscopic glass slide using a nebuliser. The sample was placed under a10X microscope objective and the spectrum recorded using a 10 s exposuretime. The excitation laser intensity was 8.4 kW/cm2. The spectra has a broad-ened peak between 630 to 850 nm which could be related to NV spectrum.However, there was no peak for NV− ZFL at 640 nm, which suggested no NVwere in the sample. Inset is an image of the region examined, which is about1 µm in diameter.
129
5 5 0 6 0 0 6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 0
2 0
4 0
6 0
8 0
1 0 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.2: PL spectra of an empty microscope glass slide. The slide was examined us-
ing a 10X microscope objective and 10 s exposure time. The excitation laserintensity was 8.4 kW/cm2.
W a v e l e n g t h ( n m )Figure 8.3: PL spectrum of 1 µm diameter microdiamond particles from Columbus
Nanoworks, deposited on a quartz slide by a nebuliser. The sample was ex-amined using a 10X microscope objective with a 10 s exposure time, ans alaser intensity of 13 kW/cm2. No NV spectrum was observed. The inset showsan image of the examined region, which is approximately 7.5 µm in diameter.
This next step was to assess whether the solvent used in the nebulisation process
affected the PL spectrum. To check this, a new sample, consisting of the same
particles, but in a powder form (not mixed with solvent), was deposited on a quartz
slide, and the PL spectrum recorded. This (Fig 8.4) showed a clear NV PL spectral
130
signature. This comparison of spectra (with and without solvent) led us to conclude
that this sample probably contained a significant level of impurities, and that, when
nebulised, only these impurities were deposited on the cover slide. The same results
were obtained for smaller sized (100 nm) diamond from the same supplier. This
suggested that it was not the nebulisation process that had been responsible for the
Figure 8.4: PL spectrum of microdiamond particles from Columbus Nanoworks with anominal diameter of 1 µm. The particles, in powder form, were placed directlyon a quartz slide. The sample was placed under a 10X microscope objective. Aspectrum acquired with 13 kW/cm2 of laser intensity and 10 s exposure time.The results show a clear NV spectrum with a sharp NV− ZPL peak at 640 nm,indicating that the powder does contain NV centres. The inset shows an imageof the examined region, which has a diameter of about 1.75 µm.
Following this, a 100 nm nanodiamond solution (from Adamas Nanotechnologies)
was prepared. The particles had been pre-suspended in DI water. A 100 µg/100 µl
sample was drop-cast onto a quartz slide, and an NV PL spectrum acquired as seen
in Fig 8.5. First, however, as a nebuliser would be used to introduce the particles to
the trap, a mixture of these particles and ethanol was first nebulised onto the quartz
slid to see if any particles containing NV could be seen. Three different clusters
of diamonds deposited on the surface were studied. The first region showed no de-
tectable NV PL. The other two regions, however, showed a clear NV spectrum as
seen as an example in Fig 8.6. As we could clearly see spectra from the samples
from Adamas Nanotechnologies, they were selected for use in the levitated experi-
ments.
131
5 5 0 6 0 0 6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 0
3 06 09 0
1 2 01 5 01 8 02 1 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.5: PL spectrum from a 100 nm nanodiamond solution from Adamas Nanotech-
nologies delivered as 1 mg/ml slurries in DI water. A sample (100 µg/100 µl)was drop-cast onto a quartz slide. The spectrum was acquired under 50X mi-croscope objective with 10 s exposure time. The excitation laser intensity was4 kW/cm2. The result was a clear NV spectrum with NV0 ZFL at 575 nmand NV− ZFL at 640 nm. The inset shows the examined region, which has adiameter about 1.5 µm.
5 5 0 6 0 0 6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 0
5
1 0
1 5
2 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.6: PL spectrum from 100 nm nanodiamond (from Adamas Nanotechnologies) de-
livered as 1 mg/ml slurries in DI water. A mixture of these particles and ethanolwas nebulised into quartz slid. The spectrum was acquired using a 50X micro-scopic objective with 10 s exposure time. The excitation laser intensity was4 kW/cm2. The result was a clear NV spectrum with NV0 ZPL at 575 nm andNV− ZPL at 640 nm. The inset shows an image of the examined region, whichhas a diameter of about 4 µm.
Commercial NV diamond available for trapping, supplied by Columbus Nanoworks
and Adamas Nanotechnologies, was investigated. The sample from Columbus
132
Nanoworks had significant level of impurities, the nebulised mixture of these sam-
ple does not contain NV diamond. As the impurities could pass through the mesh
holes in the nebuliser, but not the larger NV diamond, the impurities were the only
component of the mixture that was nebulised. However, it was clear from the PL
spectra from the second sample (from Adamas Nanotechnologies) that had a smaller
proportion of impurities compared to the sample from Columbus Nanoworks. We
therefore selected the sample supplied by Adamas Nanotechnologies for the lev-
itated experiments since we could routinely load NV diamond into the trap and
observe a clear NV spectra.
8.3 NV photoluminescence from levitated diamondControlling the NV spin of levitated diamond has some experimental complexi-
ties. These complexities derive from the difficulty of observing and monitoring
the NV PL while it is levitated. For example, the Brownian motion of optically
levitated diamond in liquid or air can move the particles out of the observation re-
gion [93, 107, 108] especially at the low trap frequency of a Paul trap. Also, most
optically levitated nanodiamond using 1064 nm laser trapping shows no NV fluores-
cence [29]. Another issue is the burning, or graphitisation, of the levitated diamond
when measurements are made at near vacuum pressure (about 1 mbar) [54]. This
occurs due to the absorption of light from the trapping laser and the reduction in
heat transfer as the surrounding gas pressure is reduced. Replacing the optical trap
with an electrical trap should resolve this last issue, and Delord et al. [31] reported
the observation of NV fluorescence from 10 µm diamond at 2×10−2 mbar. How-
ever, a 532 nm excitation laser using 700 µW of laser power was still enough to heat
up the particle and cause trap loss. In our experiment, we successfully observed NV
fluorescence from levitated nanodiamond using the 100 nm samples from Adamas
Nanotechnologies. These particles was chosen based on the tests described in the
previous section. The experimental scheme is illustrated in Fig 8.7, where a mixture
of these particles and ethanol was nebulised into the trapping region of the linear
quadrupole trap. A CCD camera was used to observe the scattered light from the
133
levitated particle, which was collected by two 50 mm achromatic lenses (AL), with-
out using any filters. We trapped using 2 kVpeak−peak on the AC drive between 2 to
9 kHz with a 31 V endcap voltage. The secular frequency at atmospheric pressure
could not be measured due to air damping. We assumed that the trapped particle
had 100 nm diameter, and from the stability condition we expected a particle charge
in the range of 1×1.6×10−19 C≤ Q≤ 62×1.6×10−19 C. The CCD camera was
then replaced with an optical fibre connected to a spectrometer (Andor Shamrock
303i spectrograph with 1200 lines/mm grating and 40 nm bandpass). The fibre was
realigned to obtain the maximum intensity of scattered green (532 nm) light before
attempting to detect NV fluorescence. Here, fluorescence was filtered by a notch
filter centred at 532 nm (NF) and a long pass filter (LP) with a cut-off wavelength
of about 610 nm. The observed NV spectrum at a laser intensity of 57 kW/cm2 is
shown in Fig 8.8. The spectrum was recorded from 610 nm to 900 nm and accumu-
lated with an exposure time of 9.12 mins. Individual scans with a 40 nm bandpass
were merged to cover the whole wavelength range. The variable voltage on the
endcap electrodes allowed us to align the particle and maximise the fluorescence
signal. It also allowed us to translate the trapped particle away from the collection
lens focal point to detect the background signal as seen in Fig 8.8. This gave us
the ability to measure the background without having to drop the particle. This will
be a useful feature, allowing us to subtract the background when the laser power
is changed or when using another laser to cool the internal temperature of NV dia-
mond coating with Yb3+:YLF crystals. Also, this feature could overcome the issue
reported by Delord et al. [31], where levitating the micro-diamond in a ring trap
shifted it from the trap centre due to the effect of the damping rate on the secular
frequency. It should be noted that during the acquisition, the strong background
signal could make detection of the NV spectrum difficult, as seen in Fig 8.9. This
noise was not apparent when the 532 nm laser was off and is most likely produced
by unwanted fluorescence from optical components along the laser optical axis.
134
Vacuum chamber
PumpVacuum
gauge
Optical fibre
Collimator Optical fibre to
APD or spectrometer
MW
NF LPF
ALAL
AC
ACGND
GND
&
MW
CCD
camera
Figure 8.7: Schematic of the levitating experiment set-up. A mixture of 100 nm nanodi-amond particles and ethanol was nebulised into the trapping region of the lin-ear quadrupole trap. An optical fibre connected to a collimator was used for532 nm laser excitation. The optical detection set-up consisted of two 50 mmachromatic lenses (AL) to collect the light. A CCD camera was used to viewthe levitated particle. For fluorescence detection, an optical fibre attached to anAPD or spectrometer was used, after filtering the fluorescence by a notch filtercentred at 533 nm (NF) and a long pass filter (LP) with cut-off wavelength ofabout 610 nm.
6 4 0 6 8 0 7 2 0 7 6 0 8 0 0 8 4 0 8 8 00
1 0
2 0
3 0
4 0
5 0
6 0
7 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )
N V s p e c r u m A f t e r r e a l i g n m e n t b y e n d c a p s O u t o f s p o t
Figure 8.8: NV spectrum of levitated diamond. The laser intensity was set at 57 kW/cm2
and spectra were accumulated twice and recorded from 610 nm to 900 nm.The exposure time was 34.16 s with a total acquisition time of 9.12 mins. Theendcap electrode voltage was used to improve the signal by shifting the particletoward the focal point of the collective lens or moving it away to obtain thebackground signal. This could be changed by tuning the endcaps voltage.
135
6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 002 04 06 08 0
1 0 01 2 01 4 01 6 01 8 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )
Figure 8.9: An NV spectrum of a levitated diamond before and after subtracting the back-ground. The green line represents the spectrum before subtracting the back-ground, the black line represents the background and the red line representsthe spectrum after subtracting the background. The laser intensity was set at57 kW/cm2 and the exposure time was 34.16 s with total acquisition time of9.12 mins. The background noise signal was approximately 71% of the signalcount.
8.4 Background noiseThe background noise must be as low as possible during the acquisition. As men-
tioned in the previous section, a background signal was found when the 532 nm laser
was turned on without trapped nanodiamond. This was probably a result of fluores-
cence from optical components, such as windows, and other trap elements along the
laser optical axis. We therefore removed the windows, but observed no consequent
reduction in the background spectrum. We then investigated other components of
the trap. The primary materials used to build the trap are stainless steel (for the AC
and endcap electrodes) and nylon (for the spacers). Using the same process as in
section 8.2, the trap components were examined by recording spectra with a Raman
microscope using 514 nm excitation. The previously-observed background spec-
tra had two broadened peaks, one between 690 nm to 750 nm, and the other from
750 nm to 825 nm (Fig 8.10), while the PL spectrum of the stainless steel electrodes
had an emission band spectrum centred at about 647 nm (Fig 8.11). As this was not
observed in the background spectrum, we conclude that the electrodes do not add to
the spectral background. Similarly, by examining the emission spectra of the nylon
136
spacers (Fig 8.12), we can see that the spectra appear at a shorter wavelength than
that of the background emission and are, therefore, also not an important source of
background in our measurements.
6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 00
2 0
4 0
6 0
8 0
1 0 0
1 2 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.10: The background spectra obtained from the levitated diamond experiment. The
laser intensity was 57 kW/cm2 and the exposure time was 34.16 s, with atotal acquisition time of 9.12 mins. The spectrum has two broadened peaksbetween 690 nm to 750 nm, and 750 nm to 825 nm.
5 5 0 6 0 0 6 5 0 7 0 0 7 5 0 8 0 0 8 5 0 9 0 0
5
1 0
1 5
2 0
2 5
3 0
Photo
lumine
scen
ce (c
ounts
/s)
W a v e l e n g t h ( n m )Figure 8.11: PL spectrum of the stainless steel electrodes used for the AC and endcap elec-
trodes, obtained using a 50X microscope objective and a 20 s exposure time.The laser intensity was 39.5 kW/cm2. The spectrum has a broadened emissionband centred at about 647 nm.
W a v e l e n g t h ( n m )Figure 8.12: PL spectrum of the nylon spacer, obtained using a 50X microscope and 20 s
exposure time. The laser intensity was 39.5 kW/cm2. The emission band wasbelow 625 nm which is not in the range of the background peaks.
We conclude that neither of the trap components, nor the windows (which are made
of quartz) are the source of the background. The background fluorescence is there-
fore probably produced by optical components such the optical fibre, notch filter,
long pass filter and achromatic lenses lying along the detection axes. These com-
ponents are made from various type of glasses, and Schartner et al. [109] reported
the observation of a background fluorescence using a 532 nm laser for a various
type of glasses, such as LLF1 which is commonly used to make optical fibres. Ac-
cording to Schartner et al., the emission band of these glasses is between 700 to
900 nm as seen in Fig 8.13. This is within the NV emission band spectrum and
the background emission range measured in our investigation. Each optical com-
ponents used could contain any of these glasses (with a different structure, depends
on the manufacturer). In our case, we could not define the glass structure for each
optical component as the manufacturer does not provide a detailed description of
the materials used.
138
Figure 8.13: PL spectra for various type of glass using a 532 nm laser with a laser powerof 25 mW, adapted from Schartner et al. [109].
Although we cannot avoid using optical lenses and filters, we can reduce the effect
of the background noise by improving the NV fluorescence signal. This weak NV
signal can be improved by replacing the two achromatic lenses that are 150 mm
away from the trapped particle (N.A. less than 0.25) with a high N.A. lens closer
to the particle. This will improve the poor light collection of the detection set-up.
This change (a lens near the trap inside the vacuum chamber) could be considered
as part of a re-engineered version of the system for future experiments.
8.5 LimitationsIn this section, we describe a few practical details that have limited observation of
the NV spectrum, which need to be taken into account to improve the experiment.
The first limitation is background emission, which contributes about 70% of the
total acquisition count. This limited the observation of the NV spectrum, especially
when performing the ODMR measurements using an APD. All components along
the optical axis of the laser were examined to verify the source of this background
noise. We found that neither the trap components nor the vacuum window flanges,
which can be changed, were the source of this background. We assumed, therefore,
that the optical filters and lenses were the source of this background fluorescence.
139
8.6 ConclusionIn this chapter, we examined the observation of NV photoluminesence from levi-
tated 100 nm diameter nanodiamond at atmospheric pressure. We also demonstrated
that the loading process using a nebuliser preferentially deposits impurities over the
diamonds. The fluorescence from these impurities increases background count, es-
pecially under vacuum as seen in the previous chapter. We measured a significant
background to the NV photoluminesence and determined that fluorescence from the
trap components did not contribute to the background. It is likely that the optical
components used in the detection system are the source of this background, as most
glasses have fluorescence emission within the range of the NV spectrum.
We conclude from this study, and from the temperature dependence study in the
previous chapter, that a purer diamond sample will be essential for carrying out
the matter-wave interferometry proposal. This sample should have a lower level of
embedded defect impurities and less amorphous carbon on the surface, to reduce
light absorption and consequent heating [110]. This will help to avoid fluorescence
quenching, especially under vacuum where the internal temperature of the diamond
increases up to 700 K. In addition, the weak NV signal, due to the low N.A. im-
posed by the trap electrodes, needs to be improved by using a different trap design.
Addressing these limits will improve the signal-to-noise ratio, which is expected to
be low for the single NV emitter nanodiamond required for the matter-wave inter-
ferometry experiment.
Chapter 9
Conclusion and future work
This thesis describes the development of an experimental platform for exploring re-
cent theoretical proposals for creating macroscopic spatial superposition using lev-
itated nano diamond containing nitrogen vacancy centres (NV) [21]. A method for
electrodynamic levitation of nanodiamond has been demonstrated as well as a sys-
tem for measuring NV fluorescence and manipulation of NV spin. The feasibility
and limitations of this system for experiments in macroscopic quantum mechanics
have been explored.
9.1 Summary of current workIn the first part of this thesis, the dynamics and trapping principles of Paul traps was
reviewed, with an outline of the properties of the nitrogen vacancy centre relevant
to matter-wave interferometry. A range of electrical trap geometries were studied,
to determine their suitability for diamond levitation. Appropriate designs, with a
deep potential well and good optical access for efficient detection of NV fluores-
cence were considered. Of several established designs, a linear quadrupole trap was
chosen. Two different traps of this type were constructed and tested at atmospheric
pressure and under vacuum. The first, a cost effective trap, was constructed from a
printed circuit board. This was easy to fabricate and had a larger numerical aperture
for enhancing signal detection. Although not eventually used in this work, it has
since found application in other levitation experiments in the laboratory. A second
trap, used in most of the work in this thesis, was of a conventional linear Paul trap
141
design with cylindrical electrodes. This trap was designed with an integrated mi-
crowave antenna for excitation of NV centres. This antenna was also one of the
Paul trap electrodes. The magnetic field strength and the energy density of the mi-
crowave field within the trap was modelled, and it was found that the average power
density at the trap centre was affected by the trap electrodes and was 26% lower
than the power density calculated for a single MW antenna electrode. We demon-
strated that the MW antenna radiated sufficient power for NV excitation suitable for
an ODMR experiment.
In the second part of this thesis, a simple and cost effective system for optically
detected magnetic resonance (ODMR) of NV diamond was developed. NV fluo-
rescence from microdiamond was used to investigate the dependence of NV photo-
luminescence as a function of temperature, laser intensity and gas pressure. It was
found that the temperature change, induced by the absorption of laser light, not only
affected the resonance frequency but also the ODMR contrast. The contrast peaked
at about 11.7±0.2 % at 380±35 K down to 3.2±0.2 % at 657±20 K. This was
observed as a resonance frequency shift below the zero field splitting (at 2.87 GHz)
with more than 42 MHz. Finally, we successful observed NV fluorescence from
levitated nanodiamond at atmospheric pressure.
9.2 Future workFor the proposed matter-wave interferometry experiment, it is essential to levitate
a NV nanodiamond under high vacuum, manipulate the spin and then readout the
spin state at the end of the protocol. In this work, we demonstrated the ability to
trap 100 nm diamond down to 4×10−3 mbar and observed NV photoluminesence
at atmospheric pressure. We also demonstrated that readout of the spin was affected
by temperature changes in the diamond due to laser induced heating at different
pressures. We found that the temperature could reach 700 K for commercial NV
diamond which will quench NV photoluminesence. This is especially important in
vacuum where heat conduction by air is low. This issue of heating has so far been
a barrier toward implementing the matter-wave experiment using optically levitated
142
NV diamond [54, 110].
In the near term, addressing this heating issue is of central importance. Using
purer diamond will help to avoid photoluminesence quenching. Morley et.al [110]
demonstrated this by using a purer milled chemical vapour deposition (CVD) nan-
odiamond. This milled CVD diamond is approximately 1000 times purer than the
commercial NV diamond used in our experiment. These pure CVD nanodiamonds
can be optically trapped with laser intensities up to 700 GW/m2 at pressures of
5 mbar and shows no heating above room temperature. Using such nanodiamonds in
a Paul trap (as opposed to an optical trap) will improve matters still further as a Paul
trap does not require such a high optical intensity. Purer diamond has yet another
advantage in that it has a longer decoherence time, T2, which determines the period
over which any experiment can be undertaken. This time could be further increased
by reducing the concentration of embedded NV centres to prevent any spin-spin
interaction between NVs [48]. A longer T2 will also help to reduce the microwave
power required for exciting NV nanodiamonds. However, the purity level required
to prevent heating of levitated NV diamond is yet to be determined. For future
experiments, it is possible to determine whether the purity is sufficiently high by
carrying out ODMR measurements under vacuum, as undertaken in chapter 7. This
provides an indication of the extent to which the optical field induces heating of pure
NV diamond at different pressures. Additional work is also required to evaluate
whether T2 for levitated NV diamond is as long as for non-levitated NV diamond.
Vamivakas et al. [111] reported that T2 is unaffected for optically levitated 100 nm
nanodiamond with a single NV centre at atmospheric pressure, and at 25 mbar.
These experiments used a 1064 nm trapping laser with a power up to 195 mW,
and a 1.8 mW exciting optical field. Vamivakas et al. found that the internal tem-
perature is slightly increased above the room temperature, with a reduction in PL
intensity, but there was no effect on T2. However, behaviour under ultra high vac-
uum conditions, at respectively higher temperatures, has not yet been explored.
Another feasible approach to overcoming the heating issue is by cooling the inter-
nal temperature of NV diamond using Yb3+:YLF nanoparticles. These nanocrystals
143
could be cooled from 300 K to 130 K by using 1031 nm light [112]. Coating the
NV diamond with Yb3+:YLF crystals could act as a heat-bath and help to control
the internal temperature. Attaching the Yb3+:YLF particles to the nanodiamond
could be achieved by using a segmented linear quadrupole trap. Benson et al. [113]
reported a method to electrostatic coupling of two levitated particles in a segment
linear quadrupole trap as seen in Fig 9.1. Two particles with opposite charges were
levitated at each end of the segmented trap, and the segmented DC electrodes were
used to push the two particles toward each other. The Coulomb force resulting from
their opposite polarity pulled them together to form a larger crystal.
AC
GND GND GND
GND GND GND
+
+
+
++ -
- -
- -
AC
GND GND GND
GND GND GND
+
+
+
++-
+ -
+ -
AC
GND GND GND
GND GND GND
+
+
+
++
-
-+
-
-
Tim
e e
volu
tio
n
Figure 9.1: Schematic of a segment linear quadrupole trap reported by Benson et al. [113].The trap consist of four parallel cylindrical rods, two of the opposing rods areconnected to an AC voltage while the segmented rods are connected to ground(GND) or DC voltage, where the + and - represent the polarity of the DC seg-ments and the particles. This result in a potential well which separates eachcharged particle in the trap. When the polarity of the segments is changed, theparticles move along the trapping axis, as a result of electrostatic attraction.The two particles at the end form a larger particle.
The next important aspect is that NV− spins do not rotate when levitated inside
a Paul trap. Any rotation of the levitated particle will change the NV orientation,
which needs to be fixed during the experiment to perform the ODMR measure-
ments. Delord et al. [42], for example, reported that the radiation pressure of a
low intensity NV optical excitation field induces rotational motion of <2 µm size
trapped NV diamond inside a Paul trap. This motion occurs using 100’s of µW of
144
exciting optical field [42]. One possible solution is to deposit magnetic material on
the NV diamond and use a magnetic field to align the particle [43]. For this ap-
proach, it is important to ensure that the process does not lead to heating as a result
of additional laser light absorption from the magnetic material.
This thesis has explored some aspects of the trapping and manipulation of NV dia-
mond spin that can be utilised for the proposed matter-wave interferometry exper-
iment. However, it has shown that there are yet many experimental and technical
issues that need to be resolved before this exciting experiment can be realised.
Bibliography
[1] C Davisson and L Germer. Diffraction of electrons by a crystal of nickel.
Physical Review, 30(6):705, 1927.
[2] G Ghirardi, A Rimini, and T Weber. Unified dynamics for microscopic and