2704

OPTICS LETTERS / Vol. 29, No. 23 / December 1, 2004

Laser frequency locking by direct measurement of detuning

A. Ratnapala, C. J. Vale, A. G. White, M. D. Harvey, N. R. Heckenberg, and H. Rubinsztein-Dunlop
School of Physical Sciences, University of Queensland, St. Lucia 4072, Australia Received May 20, 2004 We present a new method of laser frequency locking in which the feedback signal is directly proportional to the detuning from an atomic transition, even at detunings many times the natural linewidth of the transition. Our method is a form of sub-Doppler polarization spectroscopy, based on measuring two Stokes parameters (I2 and I3 ) of light transmitted through a vapor cell. It extends the linear capture range of the lock loop by as much as an order of magnitude and provides frequency discrimination equivalent to or better than those of other commonly used locking techniques. 2004 Optical Society of America OCIS codes: 300.6210, 140.3460.

Laser frequency locking is the process of controlling a lasers frequency, relative to some physical reference, by means of feedback. Various physical references, for example, cavities and atomic transitions,1 may be used, and the methods described here apply equally well in both cases. In this Letter we focus on the spectroscopy of hyperfine transitions in Rb atomic vapor, which is relevant to research with ultracold atoms. Many experiments in atomic physics require lasers to be locked to an atomic transition. Saturated absorption and polarization spectroscopy are two well-known techniques used for this purpose.1 Both rely on counterpropagating pump and probe beams to measure a single longitudinal velocity group of atoms within a Doppler-broadened prof ile. In saturation spectroscopy, the laser may be locked to the side of an absorption peak2 by comparison of a photodetector signal with a reference voltage. It is generally desirable to lock to the top of a peak, however, which one can achieve by dithering the laser frequency about a peak and performing lock-in detection.3 This provides a derivative signal with a zero crossing at the top of the peak that is convenient for locking. Polarization spectroscopy is somewhat more attractive, as it produces a narrow, sub-Doppler spectrum with a natural zero crossing at resonance4 without the need for dithering and lock-in detection.5 Pearman et al. have studied a form of polarization spectroscopy locking in detail.6 Their method corresponds to a measure of the circular birefringence (differential refractive index for orthogonal polarizations) of an atomic sample that is induced by a circularly polarized pump beam. The locking signal is linear over a range approximately equal to the power-broadened linewidth of the transition (typically 1020 MHz in Rb vapor). Our method combines polarization spectroscopy with a measure of the circular dichroism (differential absorption for two orthogonal polarizations) and greatly extends the linear region of the lock signal. Our method stems from the proposal of Harvey and White,7 which one can best understand by analyzing the Stokes parameters of the transmitted probe light. We wish to quantify the change in the polarization state of an input laser beam in terms of the anisotropic properties of the sample. As the anisotropy is def ined
0146-9592/04/232704-03$15.00/0

in terms of circular polarizations, we choose to express the total electric field in the right left (R L) circular ER basis: E EL , where ER and EL are the right and left circular electric f ield components, respectively. In this basis the (nonnormalized) Stokes parameters are I0 jER j2 1 jELj2 , (1)

I1

2jER j jELjcos f,

(2)

I2

2jER j jEL jsin f,

(3)

I3

jER j2 2 jELj2 ,

(4)

where f fR 2 fL is the phase difference between right and left circular components. Equations (1) (4) are equivalent to the familiar expressions in the horizontalvertical (H V) basis and could also be written on a diagonal antidiagonal (DA) basis. The normalized Stokes parameters, Si , are obtained from the ratio Ii I0 . We recall the physical meaning of this notation: I0 is the total irradiance and I1 , I2 , and I3 are measures of the horizontal, diagonal, and right circular polarizations, respectively.8 With this in mind it is straightforward to design detectors for each of the Stokes parameters: for I1 , a polarizing beam-splitter cube and two photodetectors; for I2 a l 2 plate and an I1 detector; and for I3 a l 4 plate and an I1 detector. Figure 1 shows such a detector schematically. I0 is given by the sum of the two photodetector signals for any of the Ii measurements above. In all cases the beam splitter separates the horizontal and vertical components, which are then measured and subtracted. The wave plates convert D A to H V to measure I2 and R L to H V to measure I3 . A measure of the phase difference between two orthogonal field components can provide an optimal feedback signal for locking a laser.7 In the circular basis, phase difference f between the left- and right-handed components is obtained from Eqs. (2) and (3). The
2004 Optical Society of America

December 1, 2004 / Vol. 29, No. 23 / OPTICS LETTERS

2705

Eq. (4). Combining expressions (5) and (6) shows that their quotient, Q x I2 x I3 x x , k (7)

Fig. 1. Detector for measuring Stokes parameters Ii , i 1, 2, 3: PD, photodector; PBS, polarizing beam splitter; li -plate, no, half-, and quarterwave plates for i 1, 2, 3, respectively; Diff. Amp., differential amplifier.

arctangent of the ratio of measured Stokes parameters I2 I1 yields a direct measure of f. In practice, calculating an arctangent in real time would require a digital signal processor. However, for small phase shifts, tan f sin f f, and I2 provides a signal proportional to f. In polarization spectroscopy6 the small-angle approximation is valid, and one obtains a signal directly proportional to f by measuring the single Stokes component, I2 . The shape of the I2 signal is given by6 I2 x ~ Dn x Da0 x , k 1 1 x2 (5)

where x is the laser detuning from resonance in units of the power-broadened transition linewidth, k is the (resonant) wave number of the laser, n x is the real part of the refractive index of the vapor, and Da0 is the difference in absorption at line center of the two circular components of the probe. This signal is dispersion shaped and free of a Doppler-broadened background (left- and right-handed components experience equal Doppler-broadened absorptions, and this is removed in the subtraction). The phase shift of relation (5) depends only on Da0 , the differential absorption at line center. This difference originates from an anisotropy induced in atoms that have interacted with the circularly polarized pump laser. Atoms are optically pumped into one of the extreme magnetic substates, say, mF 1F . Differential absorption Da of the two circular components of the probe f ield is maximized in such a case. The s 1 component experiences enhanced absorption, and the s 2 component decreased absorption, with respect to the Doppler background. For transitions between 2, mF 2, and the 5P3/2 F 0 3 the 87 Rb 5S1/2 F levels, the difference in oscillator strengths for s 1 and s 2 light is 15:1. As the Doppler-broadened absorption is the same for both components, the differential absorption prof ile is given by the Lorentzian: I3 x ~ Da x Da0 , 1 1 x2 (6)

provides a signal directly proportional to x that does not decay as 1 1 1 x2 at large detunings. In polarization locking,6 I2 is used as the error signal, and a controller provides feedback signal to the laser, which drives this to zero. I2 x is linear for small x but not at large detunings, where it approaches zero. Our locking method compensates for this nonlinearity by dividing expression (5) by expression (6), which greatly extends the linear region of the locking signal. We have implemented this method spectroscopically and constructed a simple control circuit for locking. Figure 2 is a schematic of our apparatus. We chose for our experiments the 5S1/2 F 2 ! 5P3/2 F 0 1, 2, 3 transition at 780.1 nm in 87 Rb. Light is provided by an external-cavity diode laser similar to that described in Ref. 9. Four photodetectors at the output of the optical apparatus are connected to an analog circuit that performs the necessary subtractions and division in real time. The output of this device, either polarization spectroscopy signal I2 or quotient I2 I3 , is fed into the laser lock circuit that provides feedback to the laser frequency by means of the injection current and a piezoelectric transducer. Additionally, the lock circuit has a differential input to which we can apply an offset voltage that shifts the lock point away from 0 V. Spectra were obtained as the laser frequency was scanned across the transition by application of a triangular voltage ramp to the piezoelectric transducer. Figure 3 shows the results of such a scan. The topmost trace is a saturated-absorption spectrum, followed by the raw I2 and I3 scans. The lowest trace shows quotient Q x obtained from the analog divider. The horizontal scale was obtained from the known energy-level spacings of 87 Rb in zero magnetic field. These curves show that our method greatly extends the capture range of the lock. The normalized gradients (signal-to-noise ratio) of the quotient

which is related to expression (5) by the Kramers Kronig relation.1 Expression (6) is the difference in absorption of left- and right-handed circular components (circular dichroism), which is equivalent to I3 of

Fig. 2. Experimental setup for polarization quotient locking. A small fraction of light is split off the laser output and sent into a polarization spectrometer. The I2 and I3 Stokes parameters of the probe light are measured and used to provide feedback to lock the laser. BS, beam splitter; PBS, polarizing BS; NPBS, 50 50 nonpolarizing BS.

2706

OPTICS LETTERS / Vol. 29, No. 23 / December 1, 2004

Fig. 3. Spectra for the 87 Rb 5S1/2 F 2 ! 5P3/2 F 0 1, 2, 3 transition. Top trace (SA), regular saturated absorption spectrum; spectra of I2 and I3 are shown. Bottom trace, the quotient Q x . Shaded regions indicate the monotonic capture range of I2 (upper) and Q x (lower) locking.

Fig. 4. Step responses of a laser locked by I2 and Q x locking after the dc lock point voltage is switched at t 0. Solid curves, measured error signal after switching; dashed curves, ideal response. Q x locking allows much larger jumps to be achieved without unlocking the laser.

much larger perturbations without coming out of lock, and, second, it is possible to apply large and precise frequency steps without leaving the lock slope. In laser cooling and BoseEinstein condensate experiments it is often necessary to jump the frequency of a laser such that the laser can be used for multiple tasks. For example, one can achieve a compressed magneto-optical trap or polarization cooling stage by shifting the magneto-optical traps laser frequency from a detuning of 212 MHz to 250 MHz. Ideally this would happen on a time scale of the order of 1 ms. Acousto-optical modulators can be used for this, but a simpler method is preferable. With our scheme, such frequency steps can be achieved simply by application of a dc offset voltage at the input to the lock circuit. Figure 4 shows an example of this use of dc offset voltage by plotting the step responses of a laser locked by regular I2 polarization spectroscopy and by our quotient method. With I2 locking, we can repeatably jump the laser up to 15 MHz in less than 200 ms. Larger steps may take the laser off the lock slope and cause it to relock at another zero crossing. By using quotient locking, however, we can jump from 270 MHz back to resonance in approximately the same time, making this technique versatile for application to laser cooling experiments. Similar jumps above resonance can also be achieved. In conclusion, for a Lorentzian transition, the ratio of dispersion and absorption spectra yields a measurement of detuning. Polarization spectroscopy obtains a Doppler-free dispersion spectrum by measuring nonnormalized Stokes parameter I2 . A similar measurement yields parameter I3 , the differential absorption. Combining these produces a spectroscopic measurement proportional to detuning that is ideal for use in laser locking that extends the capture region of the lock by as much as a factor of 10, which results in a robust lock and allows us to make rapid and precise jumps of laser frequency by as much as 70 MHz on either side of resonance. This research is supported by the Australian Research Council. C. J. Vales e-mail address is vale@[Link]. References
1. W. Demtrder, Laser Spectroscopy, 2nd ed. (SpringerVerlag, Berlin, 1998). 2. K. B. MacAdam, A. Steinbach, and C. Wieman, Am. J. Phys. 60, 1098 (1998). 3. A. White, IEEE J. Quantum Electron. QE-1, 349 (1965). 4. C. Wieman and T. Hnsch, Phys. Rev. Lett. 36, 1170 (1976). 5. T. Hnsch and B. Couillaud, Opt. Commun. 35, 441 (1980). 6. C. P. Pearman, C. S. Adams, S. G. Cox, P. F. Griffin, D. A. Smith, and I. G. Hughes, J. Phys. B 35, 5141 (2002). 7. M. D. Harvey and A. G. White, Opt. Commun. 221, 163 (2003). 8. M. Born and E. Wolf, Principles of Optics, 7th ed. (Cambridge U. Press, Cambridge, England, 1999). 9. A. S. Arnold, S. J. Wilson, and M. G. Boshier, Rev. Sci. Instrum. 69, 1236 (1998).

I2 I3 S2 S3 and S2 I2 I0 spectrum are the same at x 0, but Q x retains this slope at large detunings. As I3 approaches zero, Q x becomes sensitive to electronic noise on the photodetector signals. We overcome this difficulty by low-pass filtering the output of the analog divider with a 50-kHz RC f ilter. Although this f iltering limits the bandwidth of the lock loop, the linewidth and the long-term stability of lasers locked with both I2 and I2 I3 feedback signals were similar ,1 MHz , as we determined by monitoring the error signals. The linear region of the I2 locking signal extends only over 20 MHz. Our Q x signal, however, extends monotonically over a 200-MHz range, as indicated by the shaded regions in Fig. 3. At the left, it is limited by the proximity of the nearest crossover resonance; on the right, only by the voltage supplied to the analog controller. This increase in range helps in two important ways. First, it means that the laser can sustain