Scattering with pure quantum state single atoms

Ni Group1,2

1
Harvard Physics department,

2
CCB

Working on...

1 Introduction

The low energy collisional properties between atoms of different species is an important paradigm

in atomic physics, which govern phenomena multi-component quantum gases, Feshbach reso-

nances, and ultracold molecule formation. The collisions can be used to create entangled states

(Zoller 1999), or to sympathetically cool one species by another. Manipulation of the collision

can also be used to coherently create dipolar molecules by manipulating by tuning the collisional

properties using Feshbach resonances.

Common techniques for measuring the scattering properties involve starting with quantum

degenerate gases of an ensemble of atoms, and manipulating the scattering properties with Fesh-

bach resonances that enhance thermalization and atom loss. But these measurements are challeng-

ing due to the difficult in measuring atom densities.

Alternatively, single atoms can also be prepared in optical tweezers, which for the case of

1
 3 μm

Figure 1:

the same species has been used to demonstrate the Hong-Ou-Mandel effect in indistinguishable

atoms, and entanglement based on collisions. We show that a new technique for characterizing

two-atom scattering properties using the optical tweezers. The interaction between the two atoms,

in combination with their close confinement, results in a measurable modification of motional trap

frequencies. The tight trapping also enhances the interaction shift, making it straightforward to

get precise values. We demonstrate full quantum control of the motional states of the atoms and

measure these trap frequency shifts in order to characterize the two-atom the scattering properties.

In the low-energy limit, the scattering properties of the two atoms do not depend on the

fine details of the molecular potential, but are instead characterized by an s-wave scattering length

a. The scattering length depends on the spin-configuration (singlet or triplet) of the two atoms.

We measure the differential interaction shifts by performing the measuring the frequency different

different spin compositions using Raman spectroscopy. Along with measuring the binding energy

2
of the least-bound triplet state, these measurements determine the sign and magnitude of the singlet

and triplet scattering lengths.

The quantum state preparation demonstrated here is a vital step toward coherent creation of

molecules and state to state chemistry.

Collision theory When the de Broglie wavelength of the scattering atoms is too long to resolve

the fine features of the inter-atomic potential, and the molecular potential can be replaced by a

4πh̄2
Fermi pseudopotential V (r) = m
a δ (3) (r1 − r2 ) δrδ , where the amplitude is the scattering length

aS . It is a good approximation for the full molecular interaction when the atomic momentum k

is such that k|aS | 1 and becomes exact in the limit where the energy goes to zero. In optical

tweezer, the atoms are cooled nearly to the harmonic ground state, and is represented by the two-

particle wavefunction ΨN a Cs
i (r1 ) Ψj (r2 ), where i and j represents the 3D motional state, e.g. i =

(nx , ny , nz ). The scattering length can be converted to a frequency shift of the motional states by

a straight forward full diagonalization of the Hamiltonian consisting of the 2-particle harmonic

oscillator Hamiltonian and the Fermi pseudopotential.

The scattering length between the two atoms also depends on the electronic and hyperfine

spin states. The molecular potential depends on whether the two spin-1/2 atoms are in singlet

(S = 0) or triplet (S = 1) configuration, and results in the definition of both a singlet and triplet

scattering length, aS and aT . The scattering is further complicated by the presence of hyperfine

structure due to the interaction of the atom spin and nuclear moment. The single atom states are

characterized by a hyperfine state |F mF i , where the total angular momentum F = S + I consists

3
of the electronic spin S = 1/2 and the nuclear moments IN a = 3/2 and ICs = 7/2. For 23 Na, the
hf 133
hyperfine splitting between the F = 1 and F = 2 states is ∆EN a ≈ 1771.63 MHz. For Cs, the
hf
hyperfine splitting between the F = 3 and F = 4 states is ∆ECs ≈ 9192.63 MHz.

In describing 23 Na-133 Cs interactions, a two-atom channel will be labeled by |FN a , mN a ; FCs , mCs i,

with a corresponding scattering length a(FN a , mN a ; FCs , mCs ). In the elastic approximation, the

scattering length of a particular channel can be written as a(FN a , mN a ; FCs , mCs ) = PS=1 aS=1 +

PS=0 aS=0 , where PS is the probability that the channel is in either the singlet or triplet state. How-

ever, since the hyperfine interactions are much larger than the energy scales associated with the

long-range interaction −C6 /r6 , this approximation does not hold. Instead, we use a multichan-

nel quantum defect theory (MQDT) to describe the atomic interactions in each channels in terms

of the singlet s-wave scattering length, aS , the triplet s-wave scattering length, aT , the hyperfine

interactions, and the known C6 and C8 coefficients. MQDT uses a separation of the solution to

the Schrodinger equation into a long range part which is dominated by the van der Waals, and a

short range part, and has been tested extensively, including comparisons with previous numerical

calculations for the NaCs systems (Tiemann).

4
 The motional harmonic ground state for each channel is shifted by an interaction shift ∆E(FN a , mN a ; FCs , m

The differential interaction shifts are measured by using a 2-photon Raman transition to flip the

spins of the one of the atoms and measuring the frequency difference by scanning the two-photon

detuning until on-resonance. For example, in the case of flipping the Na spin, we spectroscopy

measure the frequency difference of |FN a , mN a ; FCs , mCs i → |FN0 a , m0N a ; FCs , mCs i. We then

repeat the measurement conditioned on the the second atom is not there to get the bare hyperfine

0
splitting |FN a , mN a i → |FCs , m0Cs i. The frequency difference between these two measurements

is then difference of the interactions shifts between these two states. Therefore, we will measure

the differential interaction shift between multiple channels.

To get the absolute interaction shift, we measure the binding energy of the least bound purely

triplet state. The MQDT theory provides a direct relationship between the the triplet scattering

length and the binding energy of the first bound state. The binding energy is directly measured by

preparing the atoms in the stretch state |4, 4; 2, 2i and performing two-photon spectroscopy of the

first triplet bound state.

Experiment description The experiment begins by loading a single Sodium (Na) and Cesium

(Cs) atom of Na and Cs in neighbouring tight optical tweezers inside an UHV glass cell 1, 2 . The

optical tweezers are created by focusing light through a 0.55 NA objective, and are positioned

by acoustic optic deflectors (AOD) that are driven by RF frequencies. The loading process is

stochastic, and the probability of having both a single Na and Cs atom is approximately 35%.

Before proceeding to the subsequent experiment, the atoms are imaged by collected fluorescence

5
with the objective and a CCD camera, where zero and single atoms can be distinguished with very

high fidelity. We apply Raman sideband cooling (RSC) 3, 4 to prepare the atoms in their respected

3D ground states, followed by adiabatic merging of the atoms into the same tweezer. After the

atoms are in the same trap, the effective density is nef f = 1 × 1014 cm−3 (?), temperature is 100

nK(?). Spectroscopy is performed on the two-atom state, and then the atoms are separated into

their initial tweezers and state sensitive imagining is performed again. The data is post-selected on

the initial loading condition, for example to guarantee there is both a Na and Cs atom during the

experiment.

Triplet binding energy Two-photon Raman spectroscopy is used to measure the binding energy

of the least-bound triplet state, a3 Σ (v = −1). A shown in Fig. 2A, the pump light (with Rabi

frequency Ω1 ) drives the ground harmonic state atoms to the optically excited c3 ΣΩ=1 (v = 0) state

with a single-photon detuning ∆1 . The frequency of the probe light (with Rabi frequency Ω2 ) is

scanned relative to the pump light frequency, with two-photon detuning ∆. When the two-photon

detuning ∆ is resonant to the binding energy of a molecular state, population is transferred to that

molecular state with an effective Raman Rabi frequency ΩR = Ω1 Ω2 /(2∆21 ), and simultaneous

loss of the Na and Cs atoms are observed in the experiment. As shown in Fig. 2B, survival dips

are observed for three different (Raman powers). Since the Raman light Stark shifts the transition,

we measure the binding energy at three different Raman powers in order to linearly interpolate to

the binding energy at zero Stark shift.

The atoms are initially prepared in the maximally stretched |2, 2; 4, 4i, which only contains

6
 0.8
2 mW
10 mW
A Na(3S )+Cs(6P)
0.6 4 mW

Survival
0.4
c3Σ
Potential energy

0.2
υ'=0
0
Ω2 Ω1 293 294 295 296 297
2-photon detuning (MHz)
Na(3S)+Cs(6S) 298

2-photon detuning (MHz)

295
Internuclear separation
294

293
0 2 4 6 8 10
Power (mW)

Figure 2: Description of data used to fit the binding energy: This data was taken with 1040 nm

tweezer, using tweezer as Raman light. Light is π-polarized. The power in both beams is roughly

1/2-1/2. Combining the two beams was done with three AOM’s (not using the LC waveplate).

Power for these three are (10, 4, 2) mW. The single photon frequency was 714.5 GHz, with the

c3 ΣΩ=1 (v = 0) resonance at 288696.3 GHz (fitted from light shift data). That means it is 18.2

GHz blue detuned. There is both a Stark shift from the v = 0 state and from the other states

(especially the near-threshold state), but these are power dependent and subtracted out by varying

the tweezer power. The uncertainty is most likely determined by a 1% power uncertainity for

the 2,4,10mW. I should later figure out uncertainity of the fitted value of the position. The fitted

center are (Lorentzian) 293.646(9) with FHWM 0.45(4) at 10mW, 296.008(1) with FWHM 0.07(1)

at 4mW, and 296.814(1) with 0.014(3) FWHM at 2mW. Fitting these two lines (without their

uncertainity or power uncertainities) gives 297.60(2), but the uncertainity due to power is higher,
7
so need to calculate. I should also put power unterainity on the linear graph.
the triplet S = 1 spin configuration. We look for the least bound triplet state a3 Σ(v = −1, N =

0, F = 6, mF = 6), which also only contains the triplet S = 1 configuration, and therefore

(along with the known C6 and C8 coefficient gives the triplet scattering length aT through MQDT.

The single frequency detuning is ∆ = 18.2 GHz relative to the excited state. The line-width

of the resonance is limited by the scattering of population from the least bound state off of the

excited state, and so a large detuning results in a narrow resonance and more accuracy in the

binding frequency. The π-polarized tweezer light itself is used for both branches of the two-photon

spectroscopy, with the second frequency generated with acousto-optic modulators (AOM’s) and

then recombining the light.

We perform photoassociation spectroscopy to locate the optically excited intermediate state

c3 ΣΩ=1 (v = 0). We observe the three lowest rotational states J = 1,2,3 and the hyperfine structure

within each. The frequencies fit well to E = B(J(J + 1) − Ω2 ) + E0 with E0 = 288696.? GHz,

and a rotational constant B = 1.0? GHz, (where we have used the average of the hyperfine levels

as the frequency for a specific rotational state.

When the two-photon detuning is on-resonant with the binding energy least bound state, we

observe coincidental loss of both Na and Cs due to the molecule formation, as shown in Fig. 2B

for three different Raman powers. This Raman resonance is Stark shifted by the Raman light,

and so the total power of the Raman beams is scanned in order to extrapolate a non-Stark shifted

binding energy of 297.6(1) MHz, as shown in Fig. 2C.

Extracting scattering length from 2-body interaction potentials with van der Waals long range

8
character has been done analytically? for a long range potential of the form V (r) ∝ −C6 /r6 . We

use the known C6 coefficient of a3 Σ5 , to get a triplet scattering length of aT = 30.27 a0 , where

a0 = 0.529 Å is the Bohr radius.

9
Interaction shift measurements When both Na and Cs are in the trap, the interaction between

them shifts the motional trapping frequencies. This interaction shift depends on the specific spin

configuration. Using a co-propagating optical Raman transition, we drive one of the atoms between

two Hyperfine states, and scan the 2-photon relative detuning δ to find the resonance frequency.

Since the frequency difference contains both the differential interaction shifts and the hyperfine

separation frequency, the measurement is repeated without the second atom in order to subtract

out the hyperfine splitting. This procedure is shown in Fig. 4A, where the arrows represent the

hyperfine structure, red represents driving a single atom between two hyperfine states, and blue

represents driving the atom between the hyperfine states in the presence of the second atom.

The data in Fig. 4B shows the signal from driving the transition Na(2,2)Cs(4,4)→Na(2,2)Cs(3,3)

and Cs(4,4)→Cs(3,3). The two largest peaks differ by the differential interaction shift, ∆ν1 =

32.1(2) kHz. When only Cs is present in the tweezer, we measure the “bare” transition frequency

of ω0 = 2π × 9.2 GHz. However, when Na in |2, +2i is also present, collisions between Na and

Cs lead to a differential shift of the energy of the |4, +4i and |3, +3i states due to the different

singlet/ triplet admixtures of the two Na+Cs configurations and the fact that aT 6= aS . The extra

peaks in the case with both atoms is due to the initial atoms not being fully in the ground state, and

the motional excited states have different transition frequencies.

The differential interaction shift is measured in Fig. 4C by comparing the transition frequency

Na(2,2)Cs(3,3)→Na(1,1)Cs(3,3) to Na(2,2)→Na(1,1), and we measure a differential interaction

shift of ∆ν2 = −31.8(1) kHz.

10
 0.6

Survival
0.4 Δν1

0.2

Vint 0.
-80 -40 0 40
δ/2π (kHz)

0.6

Vint Survival Δν2

0.2

0.
-80 -40 0 40 80
δ/2π (kHz)

Figure 3: Figure 3 Interaction energy shift between a single Na and Cs atom. (A) Origin of the

interaction shift. We measure the effect of the interaction shift on the ground hyperfine transition

of Cesium |F Cs , mCs
F i = |4, 4i → |3, 3i , whose “bare” value ω0 = 2π × 9.2 GHz. In the presence

↑ ↓
of a spin-polarized Na atom, the transition frequency is shifted by h∆ν1 = Vint − Vint (B) Two

measurements of the interaction shift, overlaid with Gaussian fits to the main peaks. The non-

interacting and interacting cases are in pink and blue respectively. [Top] |4, 4; 2, 2i → |3, 3; 2, 2i

gives ∆ν1 = 32.1(2) kHz [Bottom] |3, 3; 2, 2i → |3, 3; 1, 1i gives ∆ν2 = −31.8(2) kHz.

11
 We use the Fermi delta pseudo-potential formulation [cite] to numerically calculate the in-

teraction shift as a function of the total scattering length as . The matrix elements of this interaction

term are

Z Z
i,j
Vint = ((ΨN a (~r1 )ΨCs (~r2 ))i )∗ V (~r1 − ~r2 )(ΨN a (~r1 )ΨCs (~r2 ))j d3~r1 d3~r2

2πh̄2 ∂
where V (~r1 − ~r2 ) = µ
as δ (3) (~r1 − ~r2 ) ∂r r, r = |~r1 − ~r2 |, and (ΨN a (~r1 )ΨCs (~r2 ))i,j are 2-particle

spatial wavefunctions of Na+Cs, with the indices i, j denoting a particular choice of motional quan-

tum numbers (nxN a , nyN a , nzN a , nxCs , nyCs , nzCs ) (i.e., quantized excitations of the harmonic trapping

potential).

i,j
If ΨN a (~r1 ) and ΨCs (~r2 ) are regular functions, then this reduces to Vint = (2πh̄2 /µ) β i,j as ,

where the coefficients are given by β i,j = ((ΨN a (~r)ΨCs (~r))i )∗ (ΨN a (~r)ΨCs (~r))j d3~r.
R

We calculate β i,j for i and j ranging across all motional states up to a cutoff of nz < 30

in axial and nx,y < 15 in radial and a max of 0.9 to 1 MHz in energy. The cutoff eliminates

divergences caused by highly peaked density distributions of highly excited motional states and

results in an error of < 1 kHz for relevant states. (source of error?)

Fig 4B show energy shifts as a function of as (expressed in terms of the non-interacting

ground state frequency shift, (h̄/µ)as β 0,0 = as × 0.0437364 kHz/a0 ) for the 000 parity subspace.

At small as , the interactions are small and Vint ∝ as as one would expect from first order per-

turbation theory (with the slope being unity for the 3D motional ground state). However, when

the magnitude of al=0 becomes large compared to the motional state splitting, there is significant

12
 50

Vint/ħ (kHz)
0

-50

-100
- 40 - 20 0 20 40
(ħ/μ) β al=0 (kHz)
0,0

Scattering length
Figure 4: Calculated energy shift as a function of scattering length. All measurements take place

in the ground state manifold.

mixing and level repulsion results in nonlinear behaviour.

13
 From our value of aT obtained above, and using energy shifts from full diagonalization of

the motional trap eigenstates states in the presence of itnereactions and MQDT, we obtain aS =

590.7 a0 .

We first update our triplet scattering length using the latest binding energy result of 298.06(1)

MHz, from which the two-scale QDT gives us aTl=0 ≈ 30.27a.u. For the Na(2,2)Cs(4,4) state with

al=0 (2, 2; 4, 4) = aTl=0 ≈ 30.27, the diagonalization method gives ∆Eg (2, 2; 4, 4)/h ≈ 1.231 kHz,

consistent with the perturbative value of 1.324 kHz. Using the updated ∆ν1 = [∆Eg (2, 2; 4, 4) −

∆Eg (2, 2; 3, 3)]/h ≈ 32.1(2) kHz, we have ∆Eg (2, 2; 3, 3)/h ≈ −30.87kHz. From a matrix di-

agonalization calculation, this implies al=0 (2, 2; 3, 3) ≈ −297.9a.u. In comparison, a perturbation

theory would have implied al=0 (2, 2; 3, 3) ≈ −706.0a.u., which is very different. From a two-scale

MQDT, al=0 (2, 2; 3, 3) ≈ −297.9a.u. correspond to a aSl=0 ≈ 590.7a.u.. With aSl=0 ≈ 590.7a.u.

and aTl=0 ≈ 30.27a.u., MQDT predicts for the Na(1,1)Cs(3,3) channel al=0 (1, 1; 3, 3; B = 0) ≈

14.25a.u. The diagonalization method gives us ∆Eg (1, 1; 3, 3)/h ≈ 0.60kHz, in good agreement

with the perturbative value of 0.62 kHz. We thus obtain ∆ν2 = [∆Eg (2, 2; 3, 3)−∆Eg (1, 1; 3, 3)]/h ≈

−31.5kHz, in good agreement with the experimental value of -31.8(1) kHz.

The potentials in the Docenko analysis give aSl=0 = 513a.u. and aTl=0 = 33a.u., and have

C6 = 3227a.u. and C8 = 3.681 × 105 a.u.. Review how the Docenko paper go this result. One was

fitted and one uses theory.

In perturbation theory, the change of the ground trap state energy is given by ∆Eg ≈ Bg al=0 ,

14
 2h̄2 √ 1
where Bg = µ 2 +β 2 )(β 2 +β 2 )(β 2 +β 2 )
π(βx1
.
x2 y1 y2 z1 z2

We use numerical matrix diagonalization in the non-interacting basis to more precisely relate

the scattering length and ground trap state energy shift.

extracting aSl=0 . Fig 4C shows the interaction shift measurements we use to determine and

verify aSl=0 .

By comparing Na(2,2)Cs(4,4)→Na(2,2)Cs(3,3) to Cs(4,4)→Cs(3,3), we find a differential

interaction shift of ∆ν1 = 32.1(2) kHz. The shift of only the pure triplet initial state, with scat-

tering length aTl=0 , is expected to be 1.231 kHz based on Fig. 4B. Therefore ∆E(2, 2; 3, 3)/h =

−30.87 kHz. The full diagonalization implies therefore al=0 (2, 2; 3, 3) =-297.9 a0 . From MQDT

and aTl=0 obtained earlier, this gives aSl=0 = 590.7 a0 .

We also measure Na(2,2)Cs(3,3)→Na(1,1)Cs(3,3) compared to Na(2,2)→Na(1,1) to be ∆ν2 =

−31.8(2) kHz (Fig 4C. Using aSl=0 = 590.7 a0 to obtain al=0 (1, 1; 3, 3) = 14.25 a0 and a similar

procedure as above, our expected shift is 31.5 kHz, which agrees with the measured value.

fig. 4B: diagonalized energy shifts. C: 2 measurements (with fits) of interaction shift. also

indicate with dashed lines the expect potision (and strength)? based on FD and full diag

We also measure Na(2,2)Cs(3,3)→Na(1,1)Cs(3,3) compared to Na(2,2)→Na(1,1) to be ∆ν2 =

−31.8(1) kHz, which is consistent with our extracted aSl=0 . (see discussion)

15
 To go from interaction shift to scattering length, there is a prefactor which takes into account

the pair density of Na and Cs atoms. When the trapping frequencies of Na and Cs are unequal,

the prefactor has no analytic expression. We therefore rely on two independent interaction shift

measurements to eliminate the prefactor and uniquely solve for the singlet and triplet scattering

lengths. We use Cs in |3, +3i and measure the frequency of the |2, +2i → |1, +1i transition of Na.

Following the same procedure as before, we obtain an interaction shift of ∆N a = −2π × 21.9 kHz.

Fig[xyz] show the wavefunction overlaps between each state, and the non-interacting (al=0 =

0) ground state, in the parity 000 subspace, respectively. We use this to estimate the strength of

transitions beginning from the pure triplet |4, 4, 2, 2i state to any other state in the parity subspace.

This is a valid approximation because the triplet scattering length is very small.

We use the fermi delta pseudo potential formulation (validity????) to calculate the interaction

shift as a function of the total scattering length atot .

Z Z
hVint i = ΨCs (~r2 )∗ ΨN a (~r1 )∗ V (~r1 − ~r2 )ΨN a (~r1 )ΨCs (~r2 )d3~r1 d3~r2

2πh̄2 ∂
where V (~r1 − ~r2 ) = µ
atot δ (3) (~r1 − ~r2 ) ∂r r, r = |~r1 − ~r2 |, and ΨN a (~r1 ) and ΨCs (~r2 ) are spa-

tial wavefunctions of Na and Cs, respectively. If ΨN a (~r1 ) and ΨCs (~r2 ) are regular functions, then

hVint i = (2πh̄2 /µ) β atot , where β = |ΨCs (~r)ΨN a (~r)|2 d3~r. Provided we drive only carrier tran-
R

sitions and the non-paraxial effects from the tight tweezer are negligible [supplementary material],

driving the transition leaves the spatial wavefunctions of Na and Cs unchanged and ∆hVint i =

(2πh̄2 /µ) β (atot,U pper − atot,Lower ). In particular, for the transition |FCs , mF,Cs i|FN a , mF,N a i =

(|4, +4i → |3, +3i)|2, +2i, we have ∆hVint i = (2πh̄2 /µ) 0.4375 β × (aT − aS ), where the prefac-

16
tor is derived from angular momentum recoupling [supplementary material]. For |3, +3i(|2, +2i →

|1, +1i), the measured shift ∆N a = −0.2813 β × (aT − aS ). In both cases, the non-analytic ef-

fective pair density is absorbed into the pre-factor β which can be eliminated. Plugging in the

measured ∆hVint i we obtain aS = −1746.69 a0 , which is different than the value calculated in a

previous work5

2 Conclusion

We have prepared single Na and Cs in a single internal and relative motional quantum state in a

single optical tweezer. From this starting point we spectroscopically measure the singlet and triplet

scattering lengths of Na and Cs, which completely characterizes the low-energy elastic scattering

properties.

We expect this method to apply to a general class of reactants that can be optically cooled and

trapped, and immediately open the door to coherent production of designer molecules for qubits

and quantum simulation. It can also be generalized to study three body collisions in isolation (cite

bohn/ m..?), or minimal models of thermalization (cite dincao?)

17
3 Theory and Hamiltonian (SM)

For two atoms in a harmonic trap, the Hamiltonian is given by

p21 1 p2 1
H= + m1 ω12 r21 + 2 + m2 ω22 r22 + VS (|r1 − r2 |), (1)
2m1 2 2m2 2

where VS (|r1 − r2 |) is the Born-Oppenheimer (BO) molecular potential between the atoms for

when the two S = 1/2 atoms are either in the triplet (S = 1) or singlet (S = 0) configuration.

In a simplified picture, we can write this Hamiltonian in terms of the relative and center of

mass coordinates (COM), H = HCM + HR + HCM/R .

p2R 1
HR = + + µ ωR2 r2R + VS (|r|), (2)
2µ 2

which is written in terms of the reduced mass µ = (m1 m2 )/(m1 + m2 ), the relative coordinates

r = r1 − r2 , and the reduced frequency ωR = (m2 ω12 + m1 ω22 )/(m1 + m2 ).

The center-of-mass Hamiltonian is simply

p2CM 1 2
HCM = + M ωCM r2CM , (3)
2M 2

which contains the total mass M = m1 + m2 , center of mass coordinates rCM = (m1 r1 +

m2 r2 )/(m1 + m2 ), and the center-of-mass frequency ωCM = (m1 ω12 + m2 ω22 )/(m1 + m2 ).

For the case that the trapping frequencies for the two atoms are the same, then the center-of-

mass and relative coordinates are independent. Ignoring the COM coordinates reduces the problem

to just a single 3D harmonic oscillator. But when the trapping frequencies are different, there is a

18
coupling between the COM and relative coordinates given by

HR/CM = µ(ω12 − ω22 )rR rCM . (4)

q
αi
The trapping frequencies in the tweezer are proportional to ωi ∝ mi
, where the masses are

mCs = 133amu, and mN a = 23amu. The ground state polarizabilities for Na and Cs are such

that the trapping frequencies are the same at 1020 nm. At the tweezer’s wavelengths of 976 nm,

the trapping frequencies differ by 0.125.

1. Liu, L. R. et al. Building one molecule from a reservoir of two atoms. Science 360, 900–

903 (2018). URL http://science.sciencemag.org/content/360/6391/900.

http://science.sciencemag.org/content/360/6391/900.full.pdf.

2. Yu, Y. et al. Motional-ground-state cooling outside the lamb-

dicke regime. Phys. Rev. A 97, 063423 (2018). URL

https://link.aps.org/doi/10.1103/PhysRevA.97.063423.

3. Monroe, C. et al. Resolved-sideband raman cooling of a bound atom to

the 3d zero-point energy. Phys. Rev. Lett. 75, 4011–4014 (1995). URL

http://link.aps.org/doi/10.1103/PhysRevLett.75.4011.

4. Kaufman, A. M., Lester, B. J. & Regal, C. A. Cooling a Single Atom in an Op-

tical Tweezer to Its Quantum Ground State. Phys. Rev. X 2, 041014 (2012). URL

http://link.aps.org/doi/10.1103/PhysRevX.2.041014.

19
5. Docenko, O. et al. The coupling of the x 1 + and a 3 + states of the atom pair na + cs and

modelling cold collisions. Journal of Physics B: Atomic, Molecular and Optical Physics 39,

S929 (2006). URL http://stacks.iop.org/0953-4075/39/i=19/a=S08.

20