1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prl,superscriptaddress]{revtex4}
2:
3: \usepackage{graphicx}% Include figure files
4: \usepackage{dcolumn}% Align table columns on decimal point
5: \usepackage{bm}% bold math
6:
7: \def\adot{\dot{\alpha}}
8:
9: \begin{document}
10:
11: %\preprint{APS/123-QED}
12:
13: \title{Limit on the Temporal Variation of the Fine-Structure Constant\\ Using Atomic Dysprosium}
14:
15: \author{A. Cing\"{o}z}
16: \author{A. Lapierre}
17: \affiliation{Department of Physics, University of California at
18: Berkeley, Berkeley, California 94720-7300, USA}%
19: \author{A.-T. Nguyen}
20: \affiliation{Los Alamos National Laboratory, Physics Division, P-23,
21: MS-H803, Los Alamos, New Mexico 87545, USA}
22: \author{N. Leefer}
23: \affiliation{Department of Physics, University of California at
24: Berkeley, Berkeley, California 94720-7300, USA}
25: \author{D. Budker}
26: \affiliation{Department of Physics, University of California at
27: Berkeley, Berkeley, California 94720-7300, USA}%
28: \affiliation{Nuclear Science Division, Lawrence Berkeley National
29: Laboratory, Berkeley, California 94720, USA}%
30: \author{S. K. Lamoreaux}
31: \altaffiliation[Present address: ]{Department of Physics, Yale
32: University, New Haven, Connecticut 06520-8120, USA}
33: \affiliation{Los Alamos National Laboratory, Physics Division, P-23,
34: MS-H803, Los Alamos, New Mexico 87545, USA}
35: \author{J. R. Torgerson}
36: \affiliation{Los Alamos National Laboratory, Physics Division, P-23,
37: MS-H803, Los Alamos, New Mexico 87545, USA}
38:
39: \date{\today}
40: \begin{abstract}
41: Over a period of eight months, we have monitored transition
42: frequencies between nearly degenerate, opposite-parity levels in two
43: isotopes of atomic dysprosium (Dy). These transition frequencies are
44: highly sensitive to temporal variation of the fine-structure
45: constant ($\alpha$) due to relativistic corrections of large and
46: opposite sign for the opposite-parity levels. In this unique system,
47: in contrast to atomic-clock comparisons, the difference of the
48: electronic energies of the opposite-parity levels can be monitored
49: directly utilizing a radio-frequency (rf) electric-dipole transition
50: between them. Our measurements show that the frequency variation of
51: the 3.1-MHz transition in $^{163}$Dy and the 235-MHz transition in
52: $^{162}$Dy are 9.0$\pm$6.7~Hz/yr and -0.6$\pm$6.5~Hz/yr,
53: respectively. These results provide a value for the rate of
54: fractional variation of $\alpha$ of $(-2.7\pm2.6)\times
55: 10^{-15}$~yr$^{-1}$ (1 $\sigma$) without any assumptions on the
56: constancy of other fundamental constants, indicating absence of
57: significant variation at the present level of sensitivity.
58: \end{abstract}
59:
60: \pacs{06.20.Jr, 32.30.Bv}
61: %32.30.Bv: Radio-frequency, microwave, and infrared spectra
62: %06.20.Jr: Determination of fundamental constants
63: \maketitle
64:
65: Modern theories attempting to unify gravitation with the other
66: fundamental interactions have renewed interest in experimental
67: searches for temporal and spatial variation of fundamental
68: constants. These theories allow for, or even predict, such
69: variation, which would violate Einstein's Equivalence Principle
70: ~\cite{Uzan2003}. Recently, evidence for variation of $\alpha$ over
71: cosmological time scales was discovered in absorption spectra of
72: light from quasars~\cite{Webb2001, Murphy2003}. The result in
73: Ref.~\cite{Murphy2003} corresponds to
74: $\adot/\alpha=(6.40\pm1.35)\times10^{-16}$/yr assuming a linear
75: shift over $10^{10}$ years. However, more recent measurements
76: present sensitive results that are consistent with no
77: variation~\cite{Quast2004,Srianand2004}. On a geological time scale
78: of $10^9$ years, a test for variation of $\alpha$ comes from
79: analyses of fission products of a natural reactor in Oklo (Gabon).
80: There is a discrepancy between earlier
81: analyses~\cite{Damour1996,Fujii2000}, which are consistent with no
82: variation, and the analysis in Ref.~\cite{Lamoreaux2004}, where a
83: variation was reported. However, a recent full Monte Carlo
84: simulation of the reactor~\cite{Gould2006} reports a result
85: consistent with no variation at the level of $1.2\times10^{-17}$/yr.
86:
87: Observational measurements are difficult to interpret due to
88: numerous assumptions and uncontrollable systematic uncertainties.
89: Laboratory searches (see, for example, Refs.~\cite{Bize2003,
90: Marion2003, Fischer2004, Peik2004}), which probe variations on the
91: time scale of years, are easier to interpret since they are
92: repeatable, and systematic uncertainties can be studied by changing
93: experimental conditions. The best laboratory limit of
94: $|\adot/\alpha|<1.2\times10^{-15}$/yr was obtained from a comparison
95: of a Hg$^+$ optical clock to a Cs frequency
96: standard~\cite{Bize2003}. This result assumes that other constants,
97: in particular $m_e/m_p$, do not vary in time~\cite{note:memp}.
98: Interestingly, evidence for variation of this parameter on
99: astronomical time scale was recently reported in
100: Ref.~\cite{Reinhold2006}. The best limit that is independent of
101: assumptions regarding other constants was obtained by combining this
102: result with a comparison of Yb$^+$ optical clock and a Cs frequency
103: standard~\cite{Peik2004}. This measurement gives
104: $|\adot/\alpha|<2.2\times10^{-15}$/yr.
105:
106: In this Letter, we report a limit on the temporal variation of
107: $\alpha$ that is independent of other constants. This result is
108: based on observing, over a period of eight months, rf-transitions
109: between nearly degenerate, opposite-parity levels in atomic
110: dysprosium, as suggested in Ref.~\cite{Dzuba1999b}.
111:
112: The energy of an atomic level can be expressed as
113: \begin{equation}
114: E=E_0+q(\alpha^2/\alpha_0^2-1),
115: \end{equation}
116: where $E_0$ is the present-day energy, $\alpha_0$ is the present-day
117: value of the fine-structure constant, and $q$ is a level dependent
118: coefficient that determines the sensitivity to variations in
119: $\alpha$~\cite{Dzuba2003}. A theoretical calculation has found that
120: the $q$-coefficients for the two nearly degenerate, opposite-parity
121: levels in Dy (Fig.~\ref{fig:popsch}, levels $A$ and $B$) are large
122: and of opposite sign~\cite{Dzuba2003}. For the even-parity level
123: (level $A$), $q_A/hc\simeq6\times10^3$~cm$^{-1}$, while for the
124: odd-parity level (level $B$), $q_B/hc\simeq-24\times10^3$~cm$^{-1}$.
125: Since explicit uncertainties are not quoted for these numbers, they
126: are assumed to be exact in the determination of the
127: result~\cite{note:qunc}.
128:
129: Another parametrization of frequency and sensitivity factor is often
130: used for atomic-clock comparisons, in which an electronic transition
131: frequency is expressed as
132: \begin{equation}
133: \nu=R_\infty A F(\alpha),
134: \end{equation}
135: where $R_\infty$ is the Rydberg constant expressed in units of
136: frequency, $A$ is a dimensionless atomic structure factor
137: independent of $\alpha$, and $F(\alpha)$ is a factor that includes all relativistic as well as many-body effects which depend on
138: $\alpha$~\cite{Prestage1995}. The change in a transition frequency can be related to a fractional change
139: in $\alpha$ as
140: \begin{equation}
141: \frac{\dot{\nu}}{\nu}=\frac{d}{dt}\ln \nu=\left[\alpha\frac{\partial \ln F}{\partial\alpha}\right]\frac{\adot}{\alpha}.
142: \end{equation}
143: The expression $\alpha~\partial\ln F/\partial\alpha$ is the
144: sensitivity of a transition to variation in $\alpha$. Using this
145: parametrization, the q-coefficients for atomic Dy can be recast (for
146: $\alpha\approx\alpha_0$) to give $\alpha~\partial\ln
147: F/\partial\alpha = 2q/(h\nu)$, where $h\nu$ is the energy of level
148: $A$ or $B$ with respect to the ground state. This gives a
149: sensitivity of 0.6 and $-2.4$ for level $A$ and $B$, respectively.
150: These values are comparable, for example, to the sensitivity of
151: -3.2 for the transition used in the Hg$^+$ optical
152: clock~\cite{Dzuba1999}.
153:
154: \begin{figure}
155: \includegraphics{popsch.eps}
156: \caption{\label{fig:popsch} Relevant levels and transitions in
157: atomic dysprosium used to populate level $B$ and detect the
158: population of level $A$. Solid lines indicate laser excitation;
159: dashed lines indicate spontaneous emission; wavy line indicates rf
160: electric field.}
161: \end{figure}
162:
163: In atomic-clock comparisons, the observable quantity is the ratio of
164: the two frequencies being compared. A unique aspect of the Dy system
165: is that the direct observable quantity is the difference
166: $\nu_B-\nu_A$, due to the fact that an electric-dipole transition
167: can be induced between the two levels using a rf electric field. The
168: time variation of the transition frequency between levels $A$ and
169: $B$ is
170: \begin{equation}
171: \label{eq:sens}
172: \dot{\Delta\nu}=\dot{\nu}_B-\dot{\nu}_A=2\frac{q_B-q_A}{h}\frac{\adot}{\alpha}\sim-1.8\times10^{15}
173: \mbox{Hz} \frac{\adot}{\alpha}.
174: \end{equation}
175: For instance, $|\adot/\alpha|=10^{-15}$/yr implies $|\dot{\Delta
176: \nu}|\simeq2$~Hz/yr.
177:
178: There are several advantages to using the nearly degenerate levels
179: in Dy. Because most of the transition frequencies are $\sim$1~GHz or
180: smaller, direct frequency counting techniques can be used. This
181: allows for the comparison of two electronic transitions without the
182: need for optical frequency combs or transfer cavities. A preliminary
183: analysis of statistical and systematic uncertainties shows that the
184: measurement of the transition frequency and the control of the
185: systematics at a mHz level is feasible, which corresponds to an
186: ultimate sensitivity of $|\adot/\alpha|\sim10^{-18}$/yr for two
187: measurements separated by a year's time~\cite{Nguyen2004}. A mHz
188: resolution on a transition frequency of 1~GHz requires a relatively
189: modest fractional stability of $10^{-12}$ for the reference
190: frequency standard. This also means that the results are insensitive
191: to variation of the Cs reference frequency due to changes in the
192: values of fundamental constants since the experimental upper limit
193: on this relative variation rate is $\sim 10^{-15}$~yr$^{-1}$ (see,
194: for example,~\cite{Marion2003}). In addition, because the isotope
195: shifts and hyperfine splittings are on the order of the transition
196: frequencies, it is possible to work with multiple rf transitions
197: corresponding to the same electronic transition, including, in
198: particular, transitions with energy differences of opposite sign.
199: Since the variation in $\alpha$ only depends upon electronic energy
200: levels~\cite{note:hyperfine}, the variation of these rf transitions
201: should have equal magnitude but opposite sign. This correlation can
202: be used to detect and eliminate certain systematics. Currently, we
203: monitor two such transitions: the 3.1-MHz ($F=10.5\rightarrow
204: F=10.5$) transition in $^{163}$Dy and the 235-MHz ($J=10\rightarrow
205: J=10$) transition in $^{162}$Dy.
206:
207: The population and detection scheme is shown in
208: Fig.~\ref{fig:popsch}. The long-lived ($\tau_B
209: >200$~$\mu$s~\cite{Budker1994}) level $B$ is populated by
210: three transitions. The first two transitions are driven with 833-
211: and 669-nm laser light, while the third transition involves
212: spontaneous decay. Atoms are transferred to level $A$ with rf
213: electric field referenced to a Cs frequency standard. Level $A$
214: decays to the ground state in two steps. Fluorescent light at 564 nm
215: from the second decay is used for detection.
216:
217: The atomic beam is described in Ref.~\cite{Nguyen1997}. It is
218: produced by an effusive oven operating at $\sim$1500~K. In addition
219: to a multislit nozzle-array attached to the oven, two external
220: collimators are used to collimate both the atomic beam and the oven
221: light (in order to minimize the background due to scattered light
222: from the oven). The resultant atomic beam has a mean velocity of
223: 500~m/s and a full-angle divergence of $\sim$0.2~rad
224: (1/$e^2$-level).
225:
226: The laser light at 833 and 669 nm is produced by a Ti:Sapphire ring
227: laser and a ring dye laser with DCM dye, respectively. Since
228: narrow-band continuous-wave lasers are inefficient at exciting
229: atomic beams with weak collimation, an adiabatic passage technique
230: is utilized to transfer the population to level $f$ (a description
231: of this technique, as well as references to earlier work, are given
232: in Ref.~\cite{Nguyen2000}). Briefly, the beams are diverged with
233: cylindrical lenses to match the atomic-beam divergence. Due to the
234: Doppler effect, the atoms experience a frequency chirp in laser
235: detuning which transfers the population to the excited state with
236: high efficiency. The laser-light and rf-interaction region is
237: enclosed by a magnetic shield. The residual field in the interaction
238: region is $\sim$1~mG.
239:
240: The rf-generation and detection system is discussed in
241: Ref.~\cite{Cingoz2005}. The frequency modulated rf field is
242: generated by a synthesizer referenced to a commercial Cs frequency
243: standard which is compared to a second clock (Rb oscillator locked
244: to a digital cellular network signal) to monitor its stability. The
245: short term fractional stability of the clocks is $\sim10^{-11}$, and
246: it approaches $10^{-12}$ in 10 minutes of integration. The
247: modulation frequency is 10~kHz with a modulation index of 1. Since
248: the rf-transition linewidth is $\sim$20~kHz, this modulation
249: provides a fast sweep across the line shape, minimizing the effect
250: of fluctuations such as those due to laser power drifts or density
251: fluctuations of the atomic beam. The modulation is provided by the
252: reference output of a lock-in amplifier which demodulates the signal
253: from a photomultiplier tube used to detect the fluorescence. The
254: line shape of the first- and second-harmonic outputs of the lock-in
255: amplifier is, respectively, an odd function with zero crossing on
256: resonance and an even function with a maximum on resonance. In order
257: to reduce drifts further, the ratio of these two is used to measure
258: the transition frequency, which is extracted by a two-step process
259: described in detail in Ref.~\cite{Cingoz2005}.
260: %First a conversion parameter from the ratio to frequency is obtained
261: %by scanning the carrier frequency 4~kHz around the resonance and
262: %fitting a linear function to the ratio. Once the calibration is
263: %established, the rf carrier frequency is locked at a fixed value
264: %near resonance (3 074 000~Hz and 234 661 000~Hz for the two
265: %transitions), and the ratio is measured repeatedly. The average of
266: %the ratios are then converted to frequency using the conversion
267: %parameter.
268:
269: There are several imperfections that affect the stability of our
270: transition-frequency measurements. The largest effect is due to the
271: combination of the residual magnetic field with the laser-light
272: polarization imperfections. During normal operation, the
273: polarizations of both laser beams are set to linear, which should
274: lead to symmetrically populated magnetic sublevels. However,
275: residual light ellipticity leads to atomic orientation, which, in
276: the presence of the residual magnetic field, leads to asymmetric
277: broadening of the line and causes apparent shifts~\cite{Nguyen2004}.
278: In systematic studies where we deliberately used circularly
279: polarized light, shifts in the transition frequencies as large as
280: $\pm$160~Hz were observed. The linear polarizations of the laser
281: beams are determined by calcite polarizers with $10^{-5}$ extinction
282: ratio outside the vacuum chamber. However, the beams must travel
283: through optical windows and lenses utilized for adiabatic passage
284: population technique (see above) before reaching the interaction
285: region. A small amount of stress-induced birefringence on these
286: lenses was discovered by deliberate misalignment of the laser beams.
287: The amount of ellipticity depended upon where the beam sampled a
288: lens. The residual systematic uncertainties due to this effect range
289: from 2 to 5~Hz on different experimental runs with the smaller
290: uncertainty corresponding to later runs in which the lens mounts
291: were modified to relieve the stress.
292:
293: In addition to the residual magnetic field, the magnetic shielding
294: was inadvertently magnetized in the April 2006 runs during studies
295: utilizing external magnetic-field coils. Since the amount of
296: magnetization and the resultant shifts in the transition frequencies
297: were unknown at the time, systematic uncertainties have been
298: assigned by driving the magnetization to saturation and noting the
299: maximum shift induced for both field polarities. The uncertainty due
300: to this effect is 7.5~Hz. The shields were demagnetized at the
301: beginning of each subsequent run.
302:
303: Another important systematic uncertainty is due to the collisions of
304: Dy atoms with the background gas in the vacuum chamber. Collisional
305: shift rates for the rf transitions due to various gases were
306: determined in Ref.~\cite{Cingoz2005}. The pressure in the vacuum
307: chamber is $\sim 5\times 10^{-7}$~Torr when the Dy oven is off. When
308: the oven is turned on, the pressure rises to $\sim 2\times
309: 10^{-6}$~Torr, limited by H$_2$ outgassing from the oven. In
310: Ref.~\cite{Cingoz2005}, the observed shift rates for H$_2$ were
311: found to be consistent with zero. However, the pressure of other
312: gases, such as N$_2$, also increases at the level of $\sim 4\times
313: 10^{-7}$~Torr when the oven is turned on, and continually decreases
314: and stabilizes after $\sim 6$~hours of continuous oven operation.
315: The partial pressures are monitored with a residual gas analyzer
316: (RGA), and the transition frequencies are corrected for the presence
317: of N$_2$, O$_2$, H$_2$, and Ar using the shift rates measured for
318: each gas. The variation in the pressure of other gases such as
319: H$_2$O and CO$_2$ are $\sim10^{-7}$~Torr or smaller, and no
320: significant correlation between these pressure variations and the
321: transition frequency is observed. The total correction is $<2$~Hz at
322: any time during each run. The results also include small
323: uncertainties ($<0.6$~Hz) due to oven temperature variations, which
324: may indicate a possible effect of Dy intra-beam
325: collisions~\cite{Cingoz2005}.
326:
327: Finally, there are systematic effects associated with rf
328: electric-field inhomogeneities. The rf electrodes are fed by
329: twisted-pair wires at one corner. Numerical simulations show that
330: with this feed, at rf wavelengths comparable to the size of the
331: electrodes ($\sim 10$~cm), spatially varying amplitude and phase
332: inhomogeneities become substantial. These inhomogeneities can lead
333: to drifts in the transition frequency when they are combined with
334: changes in the atomic beam velocity distribution. This effect was
335: studied experimentally by deliberate detuning of the laser
336: frequencies from the optical resonances. This leads to changes in
337: the transverse velocity distribution of the excited atoms due to
338: imperfect adiabatic passage. For the two transitions considered in
339: this study, shifts of $<0.5$~Hz for 1~MHz detuning in the laser
340: frequencies were observed. To keep this effect at the 1~Hz level
341: during the runs, the laser frequencies were periodically retuned
342: ($\sim$ every 2 minutes) to the center of their respective
343: transitions. To check for slow variations in the beam velocity
344: distribution due to other changes such as collimator clogging, this
345: study was repeated several times over a two-year period, during
346: which the oven was reloaded and collimators cleaned several times.
347: No significant change in the size or dependence of the effect was
348: observed.
349:
350: All other systematics considered in Ref.~\cite{Nguyen2004} were
351: estimated to be smaller than 1 Hz for the specific conditions of the
352: current experiment. Many of these systematics, such as AC Stark
353: shifts and oven black-body radiation shifts, were studied
354: experimentally by exaggerating certain imperfections and were found
355: to be negligible.
356:
357: \begin{figure}
358: \includegraphics{paperdata.eps}
359: \caption{\label{fig:result} Measured transition frequencies for the
360: 235-MHz and 3.1-MHz transitions over an eight-month period. The data
361: have been corrected for collisional shifts. The solid lines are the
362: least-squares linear fit to the data. The dashed lines are the
363: least-squares fit to a constant function.}
364: \end{figure}
365:
366: We have measured the two rf transition frequencies in the course of
367: seven runs over eight months. Figure~\ref{fig:result} shows the
368: results of these measurements corrected for collisional shifts.
369: Uncertainties are due both to statistical and aforementioned
370: systematic uncertainties. Least-squares linear fit to the data
371: points gives slopes of $-0.6\pm 6.5$~Hz/yr and $9.0\pm6.7$~Hz/yr for
372: the 235-MHz and 3.1-MHz transitions, respectively. According to Eq.
373: (\ref{eq:sens}), these results correspond to
374: $\adot/\alpha=(-0.3\pm3.6)\times 10^{-15}$~yr$^{-1}$ for the 235-MHz
375: transition and $(-5.0\pm3.7)\times 10^{-15}$~yr$^{-1}$ for the
376: 3.1-MHz transition. As mentioned earlier, because the energy
377: differences for these two transitions are of opposite sign, these
378: two data sets can be combined to construct sum and difference
379: frequency plots. The sum frequency slope, which should be
380: insensitive to a variation in $\alpha$, is $8.1\pm9.4$~Hz/yr and
381: consistent with zero. The difference frequency, which is twice as
382: sensitive to a variation in $\alpha$ as the individual frequencies,
383: gives the final result of
384: $\adot/\alpha=(-2.7\pm2.6)\times10^{-15}$~yr$^{-1}$, consistent
385: (1~$\sigma$) with no variation at the present level of sensitivity.
386:
387: In conclusion, we have presented the first result of a direct
388: measurement of the temporal variation of $\alpha$ with atomic
389: dysprosium. This result is of comparable uncertainty to the present
390: best laboratory result that is independent of other fundamental
391: constants. However, in our case the interpretation does not require
392: comparison with different measurements to eliminate dependence on
393: other constants, and provides an alternative to measurements that
394: utilize state-of-the-art atomic optical frequency clocks. The
395: present uncertainty is dominated by systematic effects primarily due
396: to polarization imperfections coupled to the residual magnetic
397: field, collisional shifts, and rf electric-field inhomogeneities. A
398: significant improvement is expected from a new apparatus under
399: construction which will provide better control over these, as well
400: as other, systematic effects expected to be important to achieve
401: better than 1-Hz sensitivity. Ultimately, mHz-level sensitivity may
402: be achievable with this method~\cite{Nguyen2004}.
403:
404: We thank V. V. Yashchuk, D. F. Jackson Kimball, and D. English for
405: valuable discussions. This work was supported in part by the
406: University of California - Los Alamos National Laboratory CLC
407: program, NIST Precision Measurement Grant, Los Alamos National
408: Laboratory LDRD, and by grant RFP1-06-15 from the Foundational
409: Questions Institute (fqxi.org).
410:
411: \bibliography{prlpaper,footnotes}
412: \end{document}
413: