We demonstrate a simple and easy method for producing low-reflectivity surfaces that are ultra-high vacuum compatible, may be baked to high temperatures, and are easily applied even on complex surface geometries. Black cupric oxide (CuO) surfaces are chemically grown in minutes on any copper surface, allowing for low-cost, rapid prototyping, and production. The reflective properties are measured to be comparable to commercially available products for creating optically black surfaces. We describe a vacuum apparatus which uses multiple blackened copper surfaces for sensitive, low-background detection of molecules using laser-induced fluorescence.

Laser-induced fluorescence (LIF) detection is a simple yet powerful technique frequently used in gas-phase atomic and molecular spectroscopy.1–4 In many interesting situations, LIF signals are fairly small. For example, in the gas phase, rovibrational branching from molecular electronic states may lead to LIF signals of ≲1 photon/molecule. Modern experiments with atoms,5 molecules,6 and ions7 frequently aim to detect small numbers of particles—sometimes as few as one. In such cases, scattered laser light (and the noise from it) can seriously degrade the signal-to-background and signal-to-noise ratios in LIF detection.

Blackened surfaces are often used to suppress scattered light in LIF detection experiments.2,3,8 Recently, we have investigated optically black materials for stray light suppression to facilitate efficient detection of molecules via LIF (for use in experiments on laser slowing,9 cooling,10 and trapping11,12 of SrF molecules). Here, we present a method for producing ultra-high vacuum (UHV) compatible low-reflectivity surfaces that is simple to implement and very flexible in its utility. Our method, which we refer to as “copper blackening,” uses a simple chemical treatment to produce black, dendritic cupric oxide (CuO) on the surfaces of any pre-formed copper part. We describe the method and present measurements of the reflectance properties of these prepared surfaces. Measurements on blackened copper surfaces are compared to measurements on well-characterized commercial surfaces.

Blackened copper has a number of attractive properties for stray light suppression solutions. Copper is routinely machined or formed, and the surfaces may be grown in minutes on parts of any manufactured geometry. Vacuum components may be made of blackened copper, and undesired reflective surfaces may be covered with blackened copper foils or sheets formed into any desired shape. Blackened copper is suitable in UHV from cryogenic temperatures up to 500 C.13 

Copper has two common oxidization states: the red Cu2O (cuprous oxide) and the black CuO (cupric oxide). The black CuO is effectively grown on the surface of metallic copper by immersing the metal in a solution composed of 100 g of NaOH (>90% purity) and 100 g of NaClO2 (>80% purity) per liter of deionized water.14 The final step of the reaction,

Cu(OH) 2 (s) heat CuO (s) + H 2 O ( l ) ,
(1)

grows the black CuO.15 The solution’s temperature is maintained between 95 C and 100 C during the blackening process. Thermal and concentration gradients are minimized by using a magnetic stir bar in the blackening solution. Immersing the beaker holding the blackening solution in a heated outer water bath also reduces thermal gradients. The outer bath is saturated with NaCl to allow its temperature to exceed 100 C without boiling and potentially contaminating the solution.

The material used for all parts described in this paper is oxygen-free high conductivity (OFHC) copper (alloy 101), selected for its desirable UHV properties. As described below, we tested OFHC copper prepared with surfaces both as-finished, and sandblasted to increase the roughness (Fig. 1(a)). We have also performed the blackening process on alloy 110 Cu and on brass parts, which might be useful in applications where UHV is not a concern (while we have not measured the reflectance properties of these parts, they appear similarly black by eye to the 101 alloy measured in this work). Most steels and titanium are minimally corroded by the solution16 and may be used to hold or remove parts from the solution, but aluminum rapidly dissolves in NaOH.

FIG. 1.

Examples of blackened copper parts. (a) Stock (left) and sandblasted (right) copper samples before (top) and after (bottom) the blackening process. (b) Copper sheet with midsection masked from the blackening process using adhesive PTFE tape.

FIG. 1.

Examples of blackened copper parts. (a) Stock (left) and sandblasted (right) copper samples before (top) and after (bottom) the blackening process. (b) Copper sheet with midsection masked from the blackening process using adhesive PTFE tape.

Close modal

Prior to blackening, parts are cleaned in an ultrasonic bath of ∼1% Citranox in deionized water for 1 h to remove any surface contamination, followed by a rinse in a deionized water bath. Parts are then immediately placed in the blackening solution for 10 min (by eye, parts blackened for 10 min are not distinguishable from parts blackened for 5 or 15 min; under similar conditions, the oxide layer thickness was observed to saturate in 2.5 min in Ref. 17). Parts are then removed from the solution, rinsed in a fresh deionized water bath, and placed in an ultrasonic acetone bath for one hour. After a final fresh deionized water rinse, parts are dried with a gentle stream of dry N2.

CuO is known to grow on surfaces in a microstructure of fine dendrites by a number of wet-chemical15 and electrochemical methods.18,19 These dendrites effectively trap incident light,20 further reducing reflection. Care must be taken when handling the blackened parts, as the CuO dendrites are easily crushed by the pressure of touching the surface, rinsing with a squirt bottle, or drying with a high pressure of compressed gas. Crushing the dendrites is observed to result in a locally less absorptive surface. However, the dendrites are not observed to flake from the surface when crushed or when subjected to mechanical or ultrasonic vibration.

In some instances, having a part with only some areas blackened is desirable. While we find the thin CuO layer may be easily removed by mechanical means, such as sanding or machining, it is not always possible to perform such operations without crushing the dendrites in the areas where blackening is desired. We have investigated techniques for masking areas to inhibit CuO growth, and found adhesive polytetrafluoroethylene (PTFE) tape to provide good results. PTFE does not react with the blackening solution, and the acrylic adhesive leaves little visible residue; what does remain is removed in the ultrasonic acetone bath typically used in our cleaning procedure. The blackening solution creeps only a few mm under the edges of the tape during a 10-min blackening, leaving a reasonably sharp boundary (Fig. 1(b)).

When attempting to eliminate background light, the relative positions of the light source, scattering surface, and detector can significantly affect the relative intensity of scattered light. For example, surfaces typically have dramatically different reflectance properties near normal incidence and near grazing incidence. Because of this, optical reflectance is typically described in terms of the bidirectional reflectance distribution function (BRDF)21fr(θi, θr, ϕr), defined as

f r ( θ i , θ r , ϕ r ) = P r D 2 P i A cos θ r .
(2)

Here, Pr is the power reflected to the detector from incident power Pi. The geometrical quantities used to define fr are depicted in Fig. 2(a): A is the area of the detector, D is the distance from the scattering surface to the detector, θi and θr are the incident and reflected polar angles, respectively, and ϕr is the relative reflected azimuthal angle. Note that by convention, θi and θr are always in the range 0–90, while ϕr is in the range 0–180.

FIG. 2.

(a) Coordinate system used in measurements. The z-axis is the surface normal. Incident light travels in the x-z plane at angle θi to the normal. Reflected light is measured at polar angle θr and azimuthal angle ϕr on a detector of area A a distance D away. (b) Top-down view of experimental setup. The incident laser light is modulated by a chopper wheel. Rotating the turntable varies θi, while rotating the photodiode about the turntable varies θr. Under all conditions in this setup, A = 1 cm2, D = 8.2 cm, and both the incident and reflected rays lie in the x-z plane.

FIG. 2.

(a) Coordinate system used in measurements. The z-axis is the surface normal. Incident light travels in the x-z plane at angle θi to the normal. Reflected light is measured at polar angle θr and azimuthal angle ϕr on a detector of area A a distance D away. (b) Top-down view of experimental setup. The incident laser light is modulated by a chopper wheel. Rotating the turntable varies θi, while rotating the photodiode about the turntable varies θr. Under all conditions in this setup, A = 1 cm2, D = 8.2 cm, and both the incident and reflected rays lie in the x-z plane.

Close modal

If the geometric collection efficiency for the scattering source is large (i.e., the detector subtends a large solid angle) or if there are multiple scattering surfaces at different angles relative to the incident light, the directional-hemispherical reflectance (DHR)21ρ(θi), defined as

ρ ( θ i ) = f r ( θ i , θ r , ϕ r ) cos ( θ r ) sin ( θ r ) d θ r d ϕ r
(3)

(where the integration is over the reflected hemisphere), is a more useful characteristic of the surface.

For a simple relative comparison of materials, we direct a laser beam with 1/e2 diameter ≈1 mm to intersect the axis of rotation of a turntable, with the sample of interest at its center (Fig. 2(b)). The sample and detector photodiode may be rotated independently about this axis. A chopper wheel modulates the incident light power Pi at 3 kHz, and the reflected power Pr is extracted from the detector photodiode signal using a lock-in amplifier. A fraction of the beam Pm is sent to a monitor photodiode to account for source intensity fluctuations. Prior to a reflectance measurement, the signal and monitor photodiodes record Pi and Pm, respectively. During the reflectance measurement, Pi is normalized to the monitor photodiode signal.

The simplicity of this setup limits us to measurements with ϕr = 0 (forward scattering) or ϕr = 180 (back scattering). To describe these data, we define the bidirectional azimuth-independent reflection distribution function (BAIRDF) as Fr(θi, θr) = fr(θi, θr, ϕr), with θr positive (negative) for ϕr = 0 (ϕr = 180). To assign a value for the DHR from the BAIRDF at normal incidence (θi = 0), we perform the integral over ϕr in Eq. (3) by averaging over the measured values for ϕr = 0 and ϕr = 180,

ρ ( θ i ) = 0 + π / 2 F r ( θ i , θ r ) cos ( θ r ) sin ( θ r ) d θ r d ϕ r + 0 π / 2 F r ( θ i , θ r ) cos ( θ r ) sin ( θ r ) d θ r d ϕ r .
(4)

This is essentially equivalent to assuming azimuthally symmetric scattering for the case of normal incidence.

In Fig. 3, we compare Fr at normal incidence for five materials at wavelengths λ = 532 nm and 1064 nm. Values for ρ(θi = 0) are tabulated in Table I. To confirm that the calculated Fr and ρ values were reliable, ρ was measured for a near-Lambertian, high-reflectivity surface (Bright White 98). We measured ρ(θi = 0) = 0.95 at 532 nm, compared to the manufacturer-specified value 0.96 typical at 550 nm (no test data at 1064 nm were available from the manufacturer). For a commercial light-absorbing blackened surface provided on aluminum foil (Acktar Metal Velvet, AMV), we measure ρ(θi = 0) = 0.005 at 532 nm, also in fair agreement with the manufacturer’s specified typical value of 0.006.

FIG. 3.

Bidirectional azimuth-independent reflection distribution function vs reflection angle for materials investigated with λ = 532 nm (left) and λ = 1064 nm (right) at normal incidence (θi = 0). The dashed line marks a Lambertian surface with unit reflectivity.

FIG. 3.

Bidirectional azimuth-independent reflection distribution function vs reflection angle for materials investigated with λ = 532 nm (left) and λ = 1064 nm (right) at normal incidence (θi = 0). The dashed line marks a Lambertian surface with unit reflectivity.

Close modal
TABLE I.

Directional-hemispherical reflection ρ(θi = 0) for materials investigated.

Material λ = 532 nm λ = 1064 nm
Bright White 98  0.95  0.95 
Blackened Cu, stock  0.008  0.07 
Blackened Cu, sandblasted  0.008  0.06 
Black felt  0.006  0.17 
Acktar Metal Velvet  0.005  0.024 
Material λ = 532 nm λ = 1064 nm
Bright White 98  0.95  0.95 
Blackened Cu, stock  0.008  0.07 
Blackened Cu, sandblasted  0.008  0.06 
Black felt  0.006  0.17 
Acktar Metal Velvet  0.005  0.024 

In all cases investigated at normal incidence, the AMV provided the blackest surface. However, the blackened copper had only marginally higher reflectance than the AMV and proved generally comparable to standard black felt cloth. At normal incidence, no significant difference was measured between blackened copper starting with a stock finish and copper with its surface roughened by sandblasting prior to cleaning and blackening. At λ = 532 nm, blackened copper is ≈50% more reflective than black felt and ≈75% more reflective than AMV. At 1064 nm, blackened copper is ≈2 × less reflective than black felt and ≈3 × more reflective than AMV.

In Fig. 4, we plot the measured BAIRDF as a function of reflected angle θr for incident angle θi = 0 (left) and θi = 70 (right) for four common laser wavelengths λ: 405, 532, 650, and 1064 nm. The DHR for normal incidence is given in Table II. As a consistency check of our BAIRDF data, we note that by reciprocity we expect Fr(θi, θr) = Fr(θr, θi). For all cases in Fig. 4, these values agree to within 20%.

FIG. 4.

Bidirectional azimuth-independent reflection distribution function vs reflection angle for blackened stock copper (filled markers) and blackened sandblasted copper (open markers) at normal incidence (left) and 70 incidence (right).

FIG. 4.

Bidirectional azimuth-independent reflection distribution function vs reflection angle for blackened stock copper (filled markers) and blackened sandblasted copper (open markers) at normal incidence (left) and 70 incidence (right).

Close modal
TABLE II.

Directional-hemispherical reflection ρ(θi = 0) for blackened stock copper and blackened sandblasted copper vs wavelength λ.

λ (nm) Stock Sandblasted
405  0.015  0.012 
532  0.008  0.008 
650  0.010  0.013 
1064  0.07  0.06 
λ (nm) Stock Sandblasted
405  0.015  0.012 
532  0.008  0.008 
650  0.010  0.013 
1064  0.07  0.06 

At normal incidence, the blackened stock and sandblasted copper display reflectance nearly independent of θr. Both surface preparation methods produce comparable values of ρ(θi = 0) for all wavelengths tested.

Blackened copper produced from stock material was found to specularly reflect for θi ≳ 75, making it highly ineffective at reducing scattered light at near-grazing incidence. Sandblasting (using sand with typical particle diameter ∼10 μm) produces a diffuse surface over a much larger scale than the typical scale of the dendrites (∼100 nm15,20). We expected that sandblasting would tend to randomize the surface normal, and indeed no specular reflection component was observed for blackened sandblasted copper for any angle of incidence we were able to test, θi < 89. However, the removed specular reflection is accompanied by increased forward (positive θr) diffuse scatter (Fig. 4, right). For θi = 70 (an incident angle just below the onset of significant specular reflection in blackened stock copper), the stock finish has less forward scatter than the sandblasted copper, while back scatter is lower for the sandblasted copper.

Recently, we have used in-vacuum blackened copper surfaces (Fig. 5) to reduce background scattered light in experiments imaging a magneto-optical trap (MOT) of SrF molecules by LIF detection.11,12 The ultimate pressure of the vacuum chamber was measured to be ≈4 × 10−10 Torr with and without the blackened copper surfaces present. A schematic of the detection region used in Ref. 12 is shown in Fig. 6. A 14 mm 1/e2 diameter laser beam (the beam is clipped by 23 mm diameter optics external to the vacuum chamber) containing 110 mW of λ = 663 nm light is passed 6 times through the vacuum chamber (once each along three orthogonal axes and then retroreflected). Excitation by the ≈660 mW of λ = 663 nm light in the center of the chamber induces fluorescence from the trapped molecules at the same wavelength. Light is collected by a 2 in.-diameter, 150-mm-focal-length spherical singlet lens placed directly outside the vacuum chamber, followed by a 50-mm-focal-length, f/0.95 camera lens. The magnification factor at the imaging position is M = 0.45. Directly opposite to the camera is a blackened Cu screen against which the molecules are imaged. The total light detection efficiency for our system (including geometry, optical losses, and camera quantum efficiency) is η = 0.4%.11 

FIG. 5.

Examples of blackened copper parts used in LIF detection experiments. (a) CF full nipples with blackened copper inserts. (b) Blackened copper disk covers CF flange with hole in the center for optical access. (c) Blackened copper rods serves as heat (thick outer rods) and electronics conduits (thin inner rods) for in-vacuum magnet circuit boards.

FIG. 5.

Examples of blackened copper parts used in LIF detection experiments. (a) CF full nipples with blackened copper inserts. (b) Blackened copper disk covers CF flange with hole in the center for optical access. (c) Blackened copper rods serves as heat (thick outer rods) and electronics conduits (thin inner rods) for in-vacuum magnet circuit boards.

Close modal
FIG. 6.

Schematic of vacuum chamber used to detect SrF molecules in a magneto-optical trap using laser-induced fluorescence. Several blackened Cu surfaces are used to reduce scattered light. Blackened Cu heat links and electronics conduits are placed between high-power in-vacuum circuit boards. The thickness of blackened surfaces are exaggerated for clarity.

FIG. 6.

Schematic of vacuum chamber used to detect SrF molecules in a magneto-optical trap using laser-induced fluorescence. Several blackened Cu surfaces are used to reduce scattered light. Blackened Cu heat links and electronics conduits are placed between high-power in-vacuum circuit boards. The thickness of blackened surfaces are exaggerated for clarity.

Close modal

Scattering from the viewports is minimized by using high optical quality (10-5 scratch dig) glass with an anti-reflection V-coating at 663 nm. Standard vacuum nipples with blackened copper sheet inserts (Fig. 5(a)) reduce stray light in two ways. First, they reduce the power of light scattered directly from viewport surface imperfections into the LIF detection region, simply by placing the viewports further from the imaging chamber (≈10 in.) than they would be in the absence of the nipple (≈5 in.). Second, the blackened inserts increase the likelihood that light incident on the walls of the nipple (either directly or through scattering) is absorbed before reaching the imaging region. Background light is additionally reduced by baffles2 formed by blackened Cu rings with a knife-edged inner diameter machined to be 26 mm, which sit inside the imaging chamber flush with the vacuum-sealing Cu gasket.

Figure 5(b) shows a blackened copper disk for stray light absorption (not shown in Fig. 6), which covered the entire bottom of the molecule imaging chamber in Ref. 11. A hole in the center provides optical access for the two vertical passes of the 663 nm light. For the work in Ref. 12, this disk was replaced with the assembly shown in Fig. 5(c), where blackened copper rods serve as heat conduits (thick outer rods) for in-vacuum high-power circuit boards. We found that the unmodified blackened surface does not allow for good thermal contact. However, deliberately crushing the CuO dendrites on the thermally connected surface and applying a 0.004 in.-thick indium thermal interface layer (TIL) gave nearly the same thermal conductivity as obtained with a typical method for making heat links with unblackened copper (sanding the copper surface with 2000 grit sandpaper and applying an indium TIL). Using an indium TIL without crushing the dendrites gave reduced conductivity (roughly 2× less), similar to that of sanded copper without indium present.

During normal trap operation, we measure the maximum photon scattering rate per molecule to be Rsc = 2.5 × 106 s−1. Analysis is restricted to a region of interest of typically Npx = 50 × 50 pixels (3.2 × 3.2 mm area) where the majority of the LIF signal is imaged. The background scattered light signal is typically B = 1.5 × 104 e/pixel/s. For a typical te = 60 ms exposure, the noise is usually limited by shot noise in this background, at the level N B = B t e N px 1500 e . Under these conditions, the laser intensity noise, camera read noise, and camera dark current are approximately 4, 8, and 104 times smaller than NB, respectively. Integrating over the region of interest gives a signal S = ηRscte = 600 e/molecule and signal-to-noise ratio (SNR) of S/NB = 0.4/molecule for a single trap loading shot. With this system, we have observed MOTs of as few as 17 molecules,12 with SNR ≈ 110 when averaging over 300 successive shots.

We have used a procedure to produce blackened copper surfaces,14 suitable for use in UHV applications. We measure these surfaces to be only a few times more reflective than commercial UHV light-absorbing foil. While such foils are available with low-outgassing adhesives, they typically have strict temperature requirements (40 C ≲ T ≲ 150 C). Blackened copper requires no adhesive, can be baked at high temperature in vacuum, and can be applied chemically to complex surfaces as well as to flexible foils and sheets. Blackened copper has been used in our experiments for over 3 years without visible degradation of the black surface.11,12 Sandblasting prior to blackening is found to decrease diffuse back scatter, increase diffuse forward scatter, and remove specular reflection.

The advantages of vacuum-compatibility, rapid prototyping, and low cost make this technique ideally suited for scattered light suppression in experiments with LIF signals from a gas-phase source. Recently, blackened copper surfaces installed in our vacuum chamber reduced stray light noise to a level sufficient to detect a small handful of molecules via LIF.12 We also plan to use this technique to reduce scattered light background in various types of precision measurements using LIF from molecular beams,22–24 and we expect this technology will be useful in similar experiments.5–7,25

The authors thank E. R. Edwards for contributions toward the construction of the vacuum chamber. The authors acknowledge financial support from ARO and ARO (MURI). E.B.N. acknowledges funding from the NSF GRFP. N.S. acknowledges funding from Yale College Dean’s Research Fellowship.

1.
J. L.
Kinsey
,
Annu. Rev. Phys. Chem.
28
,
349
(
1977
).
2.
G.
Scoles
,
Atomic and Molecular Beam Methods
(
Oxford University Press
,
New York
,
1988
).
3.
H.
Pauly
,
Atom, Molecule, and Cluster Beams I: Basic Theory, Production, and Detection of Thermal Energy Beams
(
Springer-Verlag
,
2000
).
4.
W.
Demtröder
,
Laser Spectroscopy 2: Experimental Techniques
, 5th ed. (
Springer-Verlag
,
2015
).
5.
C.
Gross
and
I.
Bloch
, “
Microscopy of many-body states in optical lattices
,” in
Annual Review of Cold Atoms and Molecules
(
World Scientific Publishing Co.
,
2015
), Vol.
3
.
6.
W. E.
Moerner
and
D. P.
Fromm
,
Rev. Sci. Instrum.
74
,
3597
(
2003
).
7.
D.
Leibfried
,
R.
Blatt
,
C.
Monroe
, and
D.
Wineland
,
Rev. Mod. Phys.
75
,
281
(
2003
).
8.
G.
Hall
,
K.
Liu
,
M. J.
McAuliffe
,
C. F.
Giese
, and
W. R.
Gentry
,
J. Chem. Phys.
81
,
5577
(
1984
).
9.
J. F.
Barry
,
E. S.
Shuman
,
E. B.
Norrgard
, and
D.
DeMille
,
Phys. Rev. Lett.
108
,
103002
(
2012
).
10.
E. S.
Shuman
,
J. F.
Barry
, and
D.
DeMille
,
Nature
467
,
820
(
2010
).
11.
J. F.
Barry
,
D. J.
McCarron
,
E. N.
Norrgard
,
M. H.
Steinecker
, and
D.
DeMille
,
Nature
512
,
286
(
2014
).
12.
E. B.
Norrgard
,
D. J.
McCarron
,
M. H.
Steinecker
,
M. R.
Tarbutt
, and
D.
DeMille
,
Phys. Rev. Lett.
116
,
063004
(
2016
).
13.
P. D.
Kirsch
and
J. G.
Ekerdt
,
J. Appl. Phys.
90
,
4256
(
2001
).
14.
W.
Meyer
, “
Process for blackening copper or copper alloy surfaces
,” U.S. patent 2,364,993 (12 December
1944
).
15.
G.
Filipič
and
U.
Cvelbar
,
Nanotechnology
23
,
194001
(
2012
).
16.
Sandvik Materials Technology
, “
Sodium hydroxide corrosion tables
,” available online at http://smt.sandvik.com/en/materials-center/corrosion-tables/sodium-hydroxide/ (viewed 12 January 2016).
17.
M.
Lebbai
,
J.-K.
Kim
,
W.
Szeto
,
M.
Yuen
, and
P.
Tong
,
J. Electron. Mater.
32
,
558
(
2003
).
18.
W.
Shao
and
G.
Zangari
,
J. Phys. Chem. C
113
,
10097
(
2009
).
19.
Y.
Hu
,
X.
Huang
,
K.
Want
,
J.
Liu
,
J.
Jiang
,
R.
Ding
,
X.
Ji
, and
X.
Li
,
J. Solid State Chem.
183
,
662
(
2010
).
20.
Y.
Xia
,
X.
Pu
,
J.
Liu
,
J.
Liang
,
P.
Liu
,
X.
Li
, and
X.
Yu
,
J. Mater. Chem. A
2
,
6796
(
2014
).
21.
H. J.
Patrick
,
L. M.
Hanssen
,
J.
Zeng
, and
T. A.
Germer
,
Metrologia
49
,
S81
(
2012
).
22.
ACME Collaboration
,
J.
Baron
,
W. C.
Campbell
,
D.
DeMille
,
J. M.
Doyle
,
G.
Gabrielse
,
Y. V.
Gurevich
,
P. W.
Hess
,
N. R.
Hutzler
,
E.
Kirilov
,
I.
Kozyryev
,
B. R.
O’Leary
,
C. D.
Panda
,
M. F.
Parsons
,
E. S.
Petrik
,
B.
Spaun
,
A. C.
Vutha
, and
A. D.
West
,
Science
343
,
269
(
2014
).
23.
S. B.
Cahn
,
J.
Ammon
,
E.
Kirilov
,
Y. V.
Gurevich
,
D.
Murphree
,
R.
Paolino
,
D. A.
Rahmlow
,
M. G.
Kozlov
, and
D.
DeMille
,
Phys. Rev. Lett.
112
,
163002
(
2014
).
24.
L. R.
Hunter
,
S. K.
Peck
,
A. S.
Greenspon
,
S. S.
Alam
, and
D.
DeMille
,
Phys. Rev. A
85
,
012511
(
2012
).
25.
M.
Meunier
,
I.
Dutta
,
R.
Geiger
,
C.
Guerlin
,
C. L.
Garrido Alzar
, and
A.
Landragin
,
Phys. Rev. A
90
,
063633
(
2014
).