Direct observation of stopped light in a whispering-gallery-mode microresonator A. A. Savchenkov, A. B. Matsko, V. S. Ilchenko, D. Strekalov, and L. Maleki Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, California 91109-8099, USA Received 7 August 2006; published 17 August 2007 We introduce the definition of group velocity for a system with a discrete spectrum and apply it to a linear resonator. We show that a positive, negative, or zero group velocity can be obtained for light propagating in the whispering-gallery modes of a microspherical resonator. The associated group delay is practically independent of the ring-down time of the resonator. We demonstrate “stopped light” in an experiment with a fused silica microsphere. DOI: 10.1103/PhysRevA.76.023816 PACS numbers: 42.60.Da, 42.25.Bs, 42.25.Gy, 42.25.Hz The group velocity is usually introduced in a class of problems where responsivity of a material can be considered as a continuous function of frequency. For instance, prob- lems of propagation of narrowband light in systems like atomic media belong to this class 1,2. However, for the broad class of systems with discrete spectra, the common definition of the group velocity sometimes is not valid. An optical resonator is an example of such a system. We show here that the discreteness of the spectrum brings different features to the notion of the group velocity defined as the velocity of a train of optical pulses; namely, such a train can be delayed by a linear resonator for much longer than the ring-down time of the resonator. Such a delay is impossible for a single pulse interacting with a linear lossless resonator, even though linear resonators as well as their chains can introduce a significant group delay 36. This peculiarity arises when a conceptual transition is made from the framework of a distributed resonator to that of a lumped resonator. While distributed resonators possess an infinite number of modes, only a finite number of modes are usually considered for their spectral studies, and often only a single mode is retained for the sake of simplicity. Such an approximation silently transforms the distributed ob- ject to a lumped one, discarding multiple phenomena, one of which is the subject of our study. Propagation of slow light in a dispersive medium can be characterized by the propagation of a beat note envelope of two plain monochromatic electromagnetic waves E 1 = E ˜ exp-i 1 t + ik 1 z+ c.c. and E 2 = E ˜ exp-i 2 t + ik 2 z+ c.c. in the medium 2. The beat note of the waves is described by |E 1 + E 2 | 2 =2|E ˜ | 2 1+cos 1 - 2 t - k 1 - k 2 z. The veloc- ity of its propagation, V g = 1 - 2 / k 1 - k 2 , corresponds to the conventional definition of the group velocity / k if 1 2 and the wave vector k is a continuous function of frequency . Let us consider a lumped model of a resonator with a transfer function H= + i- 0 - i- 0 , 1 where is the full width at half maximum and 0 is the frequency of the resonance. A monochromatic signal with frequency acquires a phase shift argH when passing through the resonator. The resonator delays the beat note of waves E 1 and E 2 by an amount of time g 1 , 2 2/ . The maximum delay is achieved for 1 2 , and, as a general rule, g 1 , 2  1 - 2 2. As a result of the above con- ditions, there is a general belief that the group delay intro- duced by a linear resonator cannot exceed the ring-down time of the resonator. In what follows we show that this conclusion is not valid for a distributed resonator such as a microsphere resonator that supports whispering-gallery modes WGMs. The group delay in such a resonator can exceed its ring-down time sig- nificantly. This happens because a WGM resonator belongs to the class of systems with a discrete optical spectrum. Such a resonator can be described using a lumped model within each spectral line. However, the model is not valid if the light interacts with multiple modes. The usual definition of the group velocity / k does not hold in this case. For example, the expression V g = 1 - 2 / k 1 - k 2 is the only correct definition of the group velocity for a bichromatic field. A similar method should be applied to describe the propagation of a generalized optical field that has a discrete spectrum, e.g., a train of pulses, in a distributed resonator. The spectrum of the field consists of a series of arbitrarily narrowband for an arbitrarily long trainlines enveloped by the Fourier transform of an individual pulse. The number of “significant” spectral lines that are not too strongly sup- pressed by the envelope is given by essentially the duty cycle of the pulse train. This number may be just a few for a dense series of smooth e.g., Gaussianpulses. The group velocity of the train can be extremely small if the spectrum of the resonator with which the pulses interact is properly engi- neered. Let us now turn to a more formal discussion. We present the electric field inside the microsphere resonator as E = e -it + c.c., 2 where the spatial field distribution has the general form = ¯ P l m cos J l+1/2 k l,q re im / r . 3 , , r are the spherical coordinates; the indices l, m, and q determine the spatial distribution of the field, m =0,1,2,... and q = 1 , 2 , . . . are the azimuthal and radial quantization numbers, respectively, and l =0,1,2,... is the orbital mode number. , , ris the mode spatial profile, and PHYSICAL REVIEW A 76, 023816 2007 1050-2947/2007/762/0238164©2007 The American Physical Society 023816-1