next up previous
Next: Experimental Up: Generation of Weak Sub-Poissonian Previous: Generation of Weak Sub-Poissonian

Introduction

Recently there has been much interest in non-classical light both in theoretical and experimental study[1, 2]. Considerable efforts have been devoted to generate quadrature-phase squeezed states using nonlinear optical processes and to generate photon-number squeezed states or sub-Poissonian light using semiconductor devices. With these non-classical states of light, it is possible to improve the sensitivity of various measurements and the performance of communication systems beyond the shot noise limit.

The scheme utilizing semiconductor devices is especially attractive because of its simple experimental configuration and small energy consumption. A tenth noise reduction (10dB) with a laser diode (LD) was reported a few years ago[3]. Generation of ``quantum correlated twin beams"[4, 5] and realization of a ``quantum optical tap (or repeater)"[6, 7] by using light emitting diodes (LED) were also reported. In the most of the experiments, however, the light intensity was about a few milliwatts, and there is no experimental investigation of generating weak sub-Poissonian light with semiconductor devices to the best of author's knowledge.

Generation of weak sub-Poissonian light is important both from the viewpoint of fundamental research and application . This is because the lower the light intensity is, the more prominent the quantum effects become. For example, it is interesting to study nonclassical properties of sub-Poissonian light that is characterized by, e.g., the phenomenon of photon antibunching. Because the magnitude of this effect (the value of the second-order correlation function) is determined by the inverse average number of photons[10], the photon number should be small. In a photon-counting communication system using Poissonian photons (coherent light) the error probability tex2html_wrap_inline341 is given by tex2html_wrap_inline343 , where tex2html_wrap_inline345 denotes the mean photon number in one optical pulse[8]. tex2html_wrap_inline341 can be, therefore, suppressed sufficiently when the light is intense enough. However it is necessary to decrease the intensity in order to reduce the energy consumption for such application as optical interconnection. Thus in order to make the sub-Poissonian generation meaningful in a practical use, the photon number in an optical pulse must be lowered[9]. Moreover, one may obtain information about the interaction between photons and carriers in semiconductor as well as behavior of the carriers by investigating the properties of sub-Poissonian light at various intensities.

One avenue to the generation of weak sub-Poissonian light is the fabrication of a microcavity laser: by using a technique based on cavity quantum electrodynamic effect, a ``thresholdless" laser can be realized[11]. However, in the case of LD, it is known that weak optical reflection [12] and/or weak side modes [13] will degrade the amount of noise reduction. Thus care must be taken to reduce such effects. Another avenue is the use of LED. LED is intrinsically thresholdless moreover free from above-mentioned problems. It is significant, therefore, to investigate the possibility of generating weak sub-Poissonian light using conventional LED and to clarify problems associated with this scheme.

In this paper we report generation of weak sub-Poissonian light with a high-efficiency LED. The light intensity generated in our experiment is about thousand times as small as those reported in the previous experiments[4, 5, 6, 7, 16, 17]. The minimum obtainable intensity in the experiment is limited by the decrease of emission efficiency at room temperature and by the worsening of frequency response of LED at liquid nitrogen temperature. This behavior is qualitatively explained in terms of trap states.

The generation scheme of sub-Poissonian light in our experiment is based on the principle of high-impedance suppression of pump-amplitude fluctuation[14]. In this scheme, the sub-Poissonian statistics of electron current are converted directly into nonclassical photon statistics. It is convenient to use the normalized noise power (Fano factor, W) which is defined as tex2html_wrap_inline351 . If we take the emission efficiency into account, the Fano factor of the generated light, tex2html_wrap_inline353 , is related with that of the pump current, tex2html_wrap_inline355 , as [15],

equation30

where tex2html_wrap_inline357 is the normalized number fluctuation within the observation time tex2html_wrap_inline359 ( tex2html_wrap_inline361 gives the measurement frequency), and tex2html_wrap_inline363 is the probability that an injected electron will emit a photon within the delay time tex2html_wrap_inline359 . When the pump current has the Poissonian statistics ( tex2html_wrap_inline367 ), the statistics of photon is also Poissonian ( tex2html_wrap_inline369 ) regardless of the value of tex2html_wrap_inline371 . When an LED is driven by a Johnson-noise limited source and the Johnson-noise is negligible compared to quantum noise , tex2html_wrap_inline355 is effectively zero and tex2html_wrap_inline353 is given by tex2html_wrap_inline377 . The amount of noise reduction is equal to the emission efficiency. The Johnson-noise is smaller than quantum noise if tex2html_wrap_inline379 , where tex2html_wrap_inline381 is the Boltzman energy, e is the electric charge, and tex2html_wrap_inline385 is the voltage across the resistor that is series-connected to the LED[16]. At room temperature tex2html_wrap_inline387 is 52mV. Therefore it is easy to reduce the photon number fluctuation substantially if the emission efficiency is high enough.


next up previous
Next: Experimental Up: Generation of Weak Sub-Poissonian Previous: Generation of Weak Sub-Poissonian

Takuya Hirano
Fri Jun 20 15:05:14 JST 1997