Received October 9, 1996. Two-photon photoconductivity in ZnSe is used to record femtosecond autocorrelation functions. This technique requires ,100 mW of ...
March 1, 1997 / Vol. 22, No. 5 / OPTICS LETTERS
313
Femtosecond autocorrelation measurements based on two-photon photoconductivity in ZnSe W. Rudolph and M. Sheik-Bahae Department of Physics and Astronomy, University of New Mexico, Albuquerque, New Mexico 87131
A. Bernstein and L. F. Lester Center for High-Technology Materials, University of New Mexico, Albuquerque, New Mexico 87131 Received October 9, 1996 Two-photon photoconductivity in ZnSe is used to record femtosecond autocorrelation functions. This technique requires ,100 mW of average power of a typical mode-locked femtosecond Ti:sapphire laser and distinguishes itself by a dynamic range over several decades and great conversion bandwidth, permitting the sensitive correlation of pulses of a few femtoseconds. 1997 Optical Society of America
The recording of correlation functions is still the only means to gain information about the envelope and the phase of pulses shorter than a few hundred femtoseconds. A large number of nonlinear-optical processes have been exploited for single-shot and multishot recordings; see, for example, Ref. 1. The technique most commonly applied for the diagnosis of the pulses from femtosecond oscillators is second-harmonic generation, which requires a nonlinear crystal phase matched to the laser wavelength and a sensitive detector, typically a photomultiplier. Two-photon-induced photoconductivity can perform the same task by transforming the optical signal directly into an electric signal. Such a technique, based on two-photon excitation of commercially available photoreceivers and CCD arrays, was successfully applied to measure the pulse duration of a picosecond Nd:YAG laser2 and earlier to measure picosecond Nd:glass laser pulses with a CdS0.5Se0.5 photodetector.3 Photoconductivity based on two-photon ionization of gases was also used for the characterization of femtosecond pulses amplif ied in the UV.4 Here we report on the implementation of a photoconductive ZnSe switch to measure autocorrelation functions of unamplif ied femtosecond pulses. ZnSe was selected because it provides the possibility of two-photon excitation over a wavelength range of 480– 950 nm, a region where most femtosecond pulse sources operate today. The switch was fabricated as a planar metal–semiconductor–metal structure, as shown in Fig. 1. The electrode (metal) stripes, consisting of a 50-nm layer of titanium capped by a 300-nm gold layer, were deposited upon a 2-mmthick polycrystalline ZnSe substrate (II-VI, Inc.) by a standard contact photolithographic process. Two types of switch with f inger spacings d of 10 and 5 mm were examined. The dark resistance in both cases was larger than 1 GV, as measured in the typical biasing circuit of Fig. 1. To characterize the nonlinear detector we measured the electric response (i.e., the voltage drop across the load resistance RL) as a function of the average incident laser power P¯ at a constant bias voltage V 0146-9592/97/050313-03$10.00/0
38 V. Here, the load resistor RL was the 1-MV input impedance of an oscilloscope. The laser was a 120-fs, 76-MHz Ti:sapphire laser operating at 800 nm. The results are depicted in Fig. 2. We obtained curve (a) for the d 10 mm element by focusing the laser beam to a spot with a diameter slightly larger than the gap; curve ( b) was obtained with the d 5 mm element and a defocused beam to illuminate many detector f ingers simultaneously. Assuming that the conductance of the switch varies as GsP¯ d G0 1 gsP¯ d, where G0 is the dark conductance and gsP¯ d qP¯ a is the light-induced conductance, the expected voltage signal can be written as VLsP¯ d
VRL . RL 1 sG0 1 qP¯ a d21
(1)
Equation (1) fits the data points shown by solid curves in Fig. 2, yielding q s3.3 6 1.8d 3 10210 V 21 mW2a for curve (a) and a 2.16 and a 2.08 for curves (a) and ( b), respectively. From Fig. 2 it is obvious that no drastic deviation from the expected two-photon process occurs. In particular, no three-photon absorption signal was found, and the saturation behavior for incident intensities larger than 10 mW can be attributed simply to the fact that the resistance of the switch approaches RL. The larger scatter in curve (a), where tight focusing was
Fig. 1. Layout of the metal – semiconductor –metal photoconductive switch and biasing circuit. 1997 Optical Society of America
314
OPTICS LETTERS / Vol. 22, No. 5 / March 1, 1997
Fig. 2. Measured voltage as a function of the mean incident power of a 120-fs, 76-MHz Ti:sapphire laser operating at 800 nm. (a) Bias voltage V 38 V, f inger spacing d 10 mm; (b) bias voltage V 15 V, f inger spacing d 5 mm. The solid curves show theoretical fits. The inset displays an intensity autocorrelation taken at 95-mW mean power.
The smallest detectable signal was at mean powers of 10 mW. With smaller gap sizes s,10 mmd, shorter pulses s, 120 fsd, or both, this number can be reduced even further, as suggested by Eq. (2). The time constant of the detector can be controlled with the load impedance RL . At RL ø 50 V it was possible to time resolve the 76-MHz pulse train, indicating an intrinsic detector response time of better than 10 ns. Next we measured the detector response as a function of the bias voltage for two different incident powers (Fig. 3). After an initial nonlinear increase the signals show the expected linear (ohmic) behavior when the bias voltage is varied. The nonlinear part can be attributed to the existence of a small Schottky barrier at the metal –ZnSe interface. To record pulse autocorrelations, we replaced the second-harmonic-generating crystal –photomultiplier combination in a Michelson-type pulse autocorrelator by a ZnSe photoconductive switch. The intensity autocorrelation shown in the inset of Fig. 2 was taken with 95-mW incident average power. An interferometric autocorrelation with the expected 8:1 contrast ratio is depicted in Fig. 4 and shows features typical for a
used, is most likely due to a higher sensitivity to a lateral focus shift caused by insertion of attenuation f ilters in the beam path. For a quantitative estimate let us assume that the current is produced in an effective volume of d3 . The conductance (averaged over the round-trip time Trt of the laser) of the switch can be written as " # bs1 2 Rd2 P¯ 2 trec Trt em . 2 3 ¯ ¯ GsP d qP d g d2 2hnw04 tp
(2)
Here b is the two-photon absorption coeff icient, R is the ref lectivity of the substrate, trec is the carrier recombination time, m ø me is the electron mobility assuming a negligible contribution from holes, tp is the pulse duration, hn is the photon energy, and w0 is the beam waist. The expression in brackets describes the average carrier density kNl produced by two-photon absorption, where g ø 0.2 is an effective averaging factor that accounts for the temporal and spatial prof iles of the incident beam. In deriving Eq. (2) we neglected saturation (state filling), threephoton absorption, and free carrier absorption, all of which is justif ied for our illumination parameters and the resulting carrier densities of ,5 3 1017 cm23 . For b 6 3 1029 cmyW,5 me 100 cm2 yVs,6 trec 0.5 ns,7 R 0.2, w0 d, and tp 100 fs we estimate that q ø 1.3 3 1029 V 21 mW22 for the d 10 mm detector. The discrepancy (factor of ,4) between the measured and the estimated values can be attributed to uncertainties in the values for the carrier mobility, recombination time, and effective volume probed by the current. Additionally, as we shall see from the current– voltage characteristics, the metallic contacts are not perfectly ohmic, resulting in small Schottkybarrier effects that can reduce the overall gain of the detector.
Fig. 3. Measured signal as a function of the bias applied to the ZnSe two-photon photoconductive switch illuminated by a 120-fs, 76-MHz Ti:sapphire laser. The mean power used for curve ( b) was 0.45 of the mean power used for curve (a).
Fig. 4. Interferometric autocorrelation of a Ti:sapphire laser based on two-photon photoconductivity in ZnSe.
March 1, 1997 / Vol. 22, No. 5 / OPTICS LETTERS
slightly chirped pulse. In both cases the scanning was performed by a loudspeaker. The residual noise in the wings is due to the imperfect suppression of the 76-MHz pulse repetition signal. It must be noted also that the detectors used in our experiments were not wire bonded, but connections were made by contact probes. The signal-to-noise ratio (SNR) can therefore be drastically improved by proper wire bonding and shielding. Nevertheless, the demonstrated sensitivity, corresponding to SNR 1 (see the inset in Fig. 2) yields a figure of merit Ppeak 3 P¯ ø 10 mW2 , which is comparable to that in sensitive schemes that use second-harmonic-generating crystals and photomultiplier tubes. A major concern in any nonlinear autocorrelation measurement of ultrashort light pulses is the pulse broadening that is due to dispersion in the nonlinear medium. In particular, group-velocity mismatch in second-harmonic-generating crystals imposes a restriction on the usable crystal thickness. This, in turn, limits the amount of generated signal. In the two-photon conductivity arrangement, however, only a very thin surface layer (of the order of the electrode spacing d) is probed by the current. The measured dispersion effects on the pulse are therefore limited to this thickness. When the dispersion data of ZnSe are used, a length of d ø 10 mm corresponds to a dispersion length of 5-fs pulses at 800 nm. This, together with the increased sensitivity, makes two-photon conductivity an ideal choice for use in diagnosing extremely short light pulses. Higher sensitivity can be achieved with longer interaction lengths in, for example, waveguide structures.8 This higher sensitivity, however, comes only at the expense of time resolution. In conclusion, we have demonstrated two-photon conductivity in ZnSe for femtosecond pulse diagnostics.
315
The technique is sensitive and easy to implement and distinguishes itself by a broad conversion bandwidth and cost effectiveness. Research is in progress to implement arrays of detectors for single-shot autocorrelations and for correlations that do not require any mechanical scanning. Various semiconductor compounds (such as CdSe- and GaN-based compounds) are currently being investigated to increase the sensitivity and to extend the responsivity further into the near-UV spectral region. This project was supported in part by a National Science Foundation career grant (ECS-9625532) and the W. M. Keck Foundation. References 1. J.-C. Diels and W. Rudolph, Ultrashort Laser Pulse Phenomena: Fundamentals, Techniques, and Applications on a Femtosecond Time Scale (Academic, San Diego, Calif., 1996), pp. 365 –399. 2. Y. Takagi, T. Kobayashi, K. Yoshihara, and S. Imamura, Opt. Lett. 17, 658 (1992). 3. C. H. Lee, in Picosecond Optoelectronic Devices, C. H. Lee, ed. (Academic, London, 1984), pp. 156– 161. 4. S. Szatmari and F. P. Schaefer, in Ultrafast Phenomena VI, T. Yajima, K. Yoshihara, C. B. Harris, and S. Shionoya, eds. (Springer-Verlag, Berlin, 1988), pp. 82 – 86. 5. E. W. Van Stryland, M. A. Woodall, H. Vanherzeele, and M. J. Soileau, Opt. Lett. 10, 490 (1985). 6. J. I. Pankove, Optical Processes in Semiconductors (Dover, New York, 1971), p. 412. 7. E. J. Canto-Seid, D. J. Hagan, J. Young, and E. W. Van Stryland, IEEE J. Quantum Electron. 27, 2274 (1991). 8. F. R. Laughton, J. H. Marsh, D. A. Barrow, and E. L. Portnoi, IEEE J. Quantum Electron. 30, 838 (1994).
