One-chip integration of optical correlator based on slow-light devices

Transcription

1 One-chp ntegraton of optcal correlator based on slow-lght devces Shun Knugasa, Norhro Ishkura, Hroyuk Ito, Naoya Yazawa, and Toshhko Baba* Department of Electrcal and Computer Engneerng, Yokohama Natonal Unversty, 79-5 Tokwada, Hodogaya-ku, Yokohama , Japan Abstract: We propose and demonstrate an on-chp optcal correlator, n whch two types of photonc crystal slow-lght wavegudes are ntegrated and operated as an optcal delay scanner and a two-photon-absorpton photodetector. The footprnt of the devce, whch was fabrcated usng a CMOS-compatble process, was mm 2, whch s substantally smaller than that of conventonal optcal correlators wth free-space optcs. We observed optcal pulses usng ths devce and confrmed the correspondence of pulse waveforms wth those observed usng a commercal correlator when the pulse wdth was 5 7 ps. Ths devce wll acheve one-chppng of an optcal correlator and related measurement nstruments Optcal Socety of Amerca OCIS codes: ( ) Photonc crystal wavegudes; ( ) Integrated optcs devces. References and lnks 1. J. An, K. Pyun, O. Kwon, and D. E. Km, An autocorrelator based on a Fabry-Perot nterferometer, Opt. Express 21(1), (2013). 2. C. Monat, C. Grllet, M. Collns, A. Clark, J. Schroeder, C. Xong, J. L, L. O Faolan, T. F. Krauss, B. J. Eggleton, and D. J. Moss, Integrated optcal auto-correlator based on thrd-harmonc generaton n a slcon photonc crystal wavegude, Nat. Commun. 5(2), 3246 (2014). 3. H. Yu, W. Fang, X. Wu, X. Ln, L. Tong, W. Lu, A. Wang, and Y. R. Shen, Sngle nanowre optcal correlator, Nano Lett. 14(6), (2014). 4. F. R. Laughton, J. H. Marsh, D. A. Barrow, and E. L. Portno, The Two-Photon-Absorpton Semconductor Wavegude Autocorrelator, IEEE J. Quantum Electron. 30(3), (1994). 5. T. K. Lang, H. K. Tsang, I. E. Day, J. Drake, A. P. Knghts, and M. Asghar, Slcon wavegude two-photon absorpton detector at 1.5 μm wavelength for autocorrelaton measurements, Appl. Phys. Lett. 81(7), 1323 (2002). 6. Y. Takag, T. Kobayash, K. Yoshhara, and S. Imamura, Multple- and sngle-shot autocorrelator based on twophoton conductvty n semconductors, Opt. Lett. 17(9), (1992). 7. J. K. Ranka, A. L. Gaeta, A. Baltuska, M. S. Pshenchnkov, and D. A. Wersma, Autocorrelaton measurement of 6-fs pulses based on the two-photon-nduced photocurrent n a GaAsP photodode, Opt. Lett. 22(17), (1997). 8. D. J. Kane and R. Trebno, Sngle-shot measurement of the ntensty and phase of an arbtrary ultrashort pulse by usng frequency-resolved optcal gatng, Opt. Lett. 18(10), (1993). 9. A. Pasquaz, M. Peccant, Y. Park, B. E. Lttle, S. T. Chu, R. Morandott, J. Azaña, and D. J. Moss, Subpcosecond phase-senstve optcal pulse characterzaton on a chp, Nat. Photoncs 5(10), (2011). 10. P. O Shea, M. Kmmel, X. Gu, and R. Trebno, Hghly smplfed devce for ultrashort-pulse measurement, Opt. Lett. 26(12), (2001). 11. P. Gabolde and R. Trebno, Sngle-frame measurement of the complete spatotemporal ntensty and phase of ultrashort laser pulses usng wavelength-multplexed dgtal holography, J. Opt. Soc. Am. B 25(6), A25 A33 (2008). 12. P. Bowlan, U. Fuchs, R. Trebno, and U. D. Zetner, Measurng the spatotemporal electrc feld of tghtly focused ultrashort pulses wth sub-mcron spatal resoluton, Opt. Express 16(18), (2008). 13. N. Ishkura, R. Hoso, R. Hayakawa, T. Tamanuk, M. Shnkawa, and T. Baba, Photonc crystal tunable slow lght devce ntegrated wth mult-heaters, Appl. Phys. Lett. 100(22), (2012). 14. R. Hayakawa, N. Ishkura, H. C. Nguyen, and T. Baba, Hgh-speed delay tunng of slow lght n pn-dodencorporated photonc crystal wavegude, Opt. Lett. 38(15), (2013). 15. R. Hayakawa, N. Ishkura, H. C. Nguyen, and T. Baba, Two-photon absorpton photododes n S photonccrystal slow-lght wavegudes, Appl. Phys. Lett. 102(3), (2013). (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20767

2 16. M. Shnkawa, N. Ishkura, Y. Hama, K. Suzuk, and T. Baba, Nonlnear enhancement n photonc crystal slow lght wavegudes fabrcated usng CMOS-compatble process, Opt. Express 19(22), (2011). 17. T. Tamura, K. Kondo, Y. Terada, Y. Hnakura, N. Ishkura, and T. Baba, Slca-Clad Slcon Photonc Crystal Wavegudes for Wdeband Dsperson-Free Slow Lght, J. Lght. Technol. (2015, presently avalable on lne). 18. F. Shnobu, N. Ishkura, Y. Arta, T. Tamanuk, and T. Baba, Contnuously tunable slow-lght devce consstng of heater-controlled slcon mcrorng array, Opt. Express 19(14), (2011). 1. Introducton An optcal correlator s wdely used to observe short optcal pulses wth a tme duraton of less than 10 ps, whch s dffcult for optoelectronc devces. It fundamentally comprses an optcal delay scanner and a nonlnear photodetector. Because a movable mrror s usually used as the delay scanner, commercal optcal correlators are box-szed devces. Thus far, downszng of optcal correlators has been studed usng a Fabry Perot resonator as a delay scanner, a photonc crystal wavegude (PCW) or wre wavegude as a second harmonc generator (SHG) n front of the photodetector [1 3], and two-photon-absorpton photodetector (TPA-PD) nstead of SHG [4 7]. Furthermore, some methods other than optcal correlators exst for measurng short optcal pulses, whch nclude frequency-resolved optcal gatng (FROG) [8, 9], smplfed FROG [10], dgtal holography [11], and near-feld scannng mcroscopy [12], among others. However, all these methods nvolve large systems wth free-space optcs. Thus far, no precedents exst for demonstratng on-chp optcal correlators and short pulse measurement nstruments. S photoncs s a fascnatng platform that enables not only large-scale photonc ntegraton on a slcon-on-nsulator (SOI) wafer wth hgh unformty and reproducblty but also low-cost mass producton va a CMOS-compatble process. We have studed an S PCW fabrcated by such a process and ts slow-lght generaton, demonstratng tunable delays [13, 14] and a TPA-PD [15] at telecom wavelengths. In ths study, we propose and demonstrate an on-chp optcal correlator n whch the two basc components are ntegrated monolthcally. Fg. 1. Schematc of the on-chp optcal correlator. 2. Desgn Fgure 1 shows a schematc of the devce structure. S-wre wavegudes termnated by spotsze converters at end facets of the chp were used for optcal wrng. To avod nonlnear nduced absorpton and pulse dstorton n the wre wavegudes, straght sectons of the optcal wrng was wdened through a taper so that the optcal ntensty was suppressed. The nput optcal pulse was dvded nto a reference branch and a delay scanner branch through a mult-mode nterferometer (MMI) 1 2 branch. A heater was ntegrated nto the reference branch for phase tunng, where the change of the delay by the heatng was neglgble. In a delay scanner based on a lattce-shfted photonc crystal wavegude (LSPCW) [13 16], the delay s scanned by ntegrated heaters ntegrated besde the LSPCW. Ths delay scanner s followed by a LSPCW dsperson controller operated by other heaters. The operaton of the delay scanner and dsperson controller can be observed by tappng off the lght va an MMI 10% tap placed adjacent to the LSPCW and extractng t through montor port #1 for (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20768

3 observaton usng an external cross-correlator. Eventually, pulses n the two branches are joned through a 2 1 MMI combner, where the n-phase condton between the two branches s mantaned by the heater of the reference branch; the pulses are subsequently launched nto the LSPCW wth a pn juncton, whch functons as a TPA-PD. The TPA effcency s enhanced by low-dsperson slow lght. A heater was also placed on the LSPCW to adjust the slow-lght band. We obtaned an auto-correlaton waveform by measurng the photocurrent from the TPA-PD as a functon of the delay. Now, we obtan the expresson for the photocurrent. The temporal electrc feld of the pulse n each branch s represented as j( ωt+ ϕ ) () () ( 1,2) E t = I t e = (1) where ω and φ are the center frequency and phase, respectvely, and = 1 and 2 correspond to reference and delay scanner branch, respectvely. I (t) s the envelope ntensty of the pulse, whch s gven by the ntensty of the ncdent pulse, I 0 (t), and attenuaton n each branch, L, as () 0 () 2 I t = LI t (2) Here, 1/2 shows the frst dvson of the pulse. The photocurrent as a functon of the delay τ s j ωt+ ϕ1 ( ) () ( ) = C ( τ ) + C ( τ ) cos( ωτ +Δ ϕ ) + C ( τ ) cos 2( ωτ +Δϕ ) ( ) j{ ωτ ( t) + ϕ2} 1 2 1() 2( ) τ = E t + E τ t dt = I t e + I τ t e dt where Δφ = φ 1 φ (3) C0( τ) = L1I0 ( t) L2I0 ( τ t) 4LL 1 2I0( t) I0( τ t) dt 1 C1( τ) = LLI 1 2 0() t I0( t) LI 1 0() t LI 2 0( t) dt 4 τ τ + 1 C2( τ) = LL 1 2I0() t I0( τ t) dt 8 (4) and 1 2 n Eq. (3) shows the feld fracton at the combner. As descrbed later, we fxed the delay τ n each measurement and adjusted the Δφ wth the phase tuner n the reference branch so that cos(ωτ + Δφ) and cos2(ωτ + Δφ) became unty and (τ) took the maxmum value. The frst and second terms n the ntegral of C 0 (τ) denote the background components whch are ndependent of τ. Other terms depend on τ and form a pulse peak. Therefore, the peak to background rato at τ = 0 s Peak : Background = L + L + 6LL + 4 LL ( L + L ) : L + L (5) In the lossless case,.e., L 1 = L 2 = 1, ths rato becomes 8: 1. If τ s scanned so fast that the detector cannot respond, the terms of the cosne functons n Eq. (3) vansh and the rato becomes 3: 1. Wth a loss, ths rato changes, as shown later. 3. Fabrcaton Fgure 2 shows the devce fabrcated on an 8-nch SOI wafer wth a 210-nm-thck S layer and a 2-µm-thck slca BOX layer usng a CMOS-compatble process whose mnmum feature sze s 180 nm. The spot sze converter has a 200-µm-long S-wre nverse taper wth a (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20769

4 180-nm-tp wdth, whch s cladded by a slca wavegude of 4 4 µm2 cross-secton [16]. The nserton loss between the spotsze converter and lensed fber wth a spot dameter of 3 µm was approxmately 2 db. The S-wre wavegudes for the optcal wrng were 400 nm wde. They were partally wdened to 4 µm va a 50-µm-long taper n the long straght secton to avod the nonlnear effects, as prevously dscussed. Even n the wdened wavegude, the exctaton of unnecessary hgher-order modes was neglgble. The reference branch comprsed four wre wavegudes (total length of 99.6 µm and calculated group ndex ng = 4.15), sx tapers (300 µm, average ng = 3.93), and three wdened wavegudes (2224 µm, ng = 3.71). The total group delay was expected to be 34 ps. The resstance of a TN heater placed on the longest wdened wavegude was 2.5 kω. The measured excess loss of the 1 2 branch Fg. 2. Fabrcated devce and detals of the MMI branch, 10% tap, and LSPCW. Fgures n the branch and 10% tap are n unts of mcrons. The doped regon n the rght-upper pcture of TPA-PD s colored schematcally. descrbed n Fg. 2 was 0.23 db. The measured tap rato and excess loss of the tap also descrbed n Fg. 2 were 10% and 1.7 db, respectvely, at λ = 1550 nm. They were suffcent for measurng waveforms usng the external correlator. We used the thrd-row-shfted LSPCW to generate low-dsperson slow lght. The 250µm-long ar-brdge LSPCW wth a lattce constant a = 450 nm, hole dameter 2r = 255 nm, and lattce shft s = 90 nm was used as the delay scanner and dsperson controller. A bow-te heater of 230-Ω resstance was employed n the delay scanner, amng to form a temperature slope and ndex chrp. Here three types of slow-lght bands were produced n the LSPCW, as shown n Fg. 3: (1) a hghly dspersve band on the short-wavelength sde; (2) a slow-lght band n whch tunable delay and dsperson compensaton are avalable va the ndex chrp; and (3) a low-dsperson slow-lght band on the long-wavelength sde. We frst planned to tune the delay usng band (2). In ths study, however, the desrable tunable slow lght was not obtaned because of some damages n the ar-brdge LSPCW, whch mght be caused by the chemcal echng prosess of the BOX layer beneath the S layer. Therefore, we set the center wavelength of the nput pulse at band (1). Here the delay τ was tuned from 3 ps to 21 ps # (C) 2015 OSA Receved 11 Jun 2015; revsed 26 Jul 2015; accepted 28 Jul 2015; publshed 30 Jul Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20770

5 smply by the spectral shft due to the heatng (red and orange lnes n Fg. 3). The loss caused by the delay, whch was evaluated through the tap, was as hgh as db (3.0 db for τ < 11 ps and 0.44τ 1.8 db for τ > 11 ps, 3.5 db at the center) due to the damage. The dsperson slope of band (1) was measured to be ps/nm. When the full-wdth at halfmaxmum (FWHM) of the pulse duraton s 5 ps, the correspondng spectral FWHM n wavelength becomes 1.3 nm, whch suggests that the pulse could be broadened to >10 ps. Actually t was compensated n the dsperson controller. Because heaters n the dsperson controller were not operated n ths study, the wavelengths of the bands were shorter than those n the delay scanner under heatng. Ths means that the band (1) n the delay scanner overlaps wth band (2) n the dsperson controller and the dsperson s compensated. The group delay of the dsperson controller was approxmately 8 ps, and that at wre wavegudes, ncludng the wdened ones, was 14 ps, resultng n a total delay of ps n ths branch. Ths delay s comparable to that n the reference branch. The 300-µm-long slca-cladded LSPCW wth a = 400 nm, 2r = 220 nm, and s = 130 nm were used as the TPA-PD. Here we Fg. 3. Group delay spectrum showng three dfferent bands (1) (3). Black lne shows no heatng condton, whle red and orange show heatng condtons wth heatng powers of 293 and 835 mw, respectvely. used band (3) so that the TPA effcency was enhanced wth the pulse shape sustaned. The resstance of the heater above ths LSPCW s 420 Ω. The footprnt ncludng all these components falls nsde mm Auto-correlaton waveform In the measurement of the correlaton waveform, the temporal FWHM of the ncdent pulse was 3 10 ps, the correspondng spectral FWHM was nm, respectvely, the center wavelength was nm, and the repetton frequency was 40 MHz. The upper panel of Fg. 4(a) shows the delay of the 5-ps pulse and change of the cross-correlaton waveform for heater power P H appled to the delay scanner; ths pulse and waveform were observed through montor port #1. The tunng range was 18 ps (τ = 3 21 ps), and the relatve delay to the reference branch became zero at P H = 230 mw. Thus, ths tunng range was sutable for observng the 5-ps pulse. The lower panel of Fg. 4(a) shows the change of the temporal FWHM. The dsperson appeared at low heatng powers, but t fnally converged to approxmately 5 ps because of the dsperson compensaton, as prevously mentoned. The FWHM slghtly smaller than the ncdent one mght be due to the slght pre-chrpng n the ncdent pulse; we observed that the spectral broadenng caused by some nonlnear effects n the LSPCW was neglgble. Fgure 4(b) shows the responsvty characterstcs of the TPA-PD measured by nsertng the pulse through montor port #2 drectly from outsde and applyng a 3 V bas. Here the pulse peak power P s derved from the tme-averaged power P av wth consderng the duty rato of the pulse. The photocurrent at P < 2 W s approxmated as I [A] = (P [W]) 2, (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20771

6 ndcatng that the current s generated by TPA. The saturaton of the photocurrent at P > 3 W mght be caused by the TPA n the optcal wrng before the TPA-PD [15]. The dark current of the TPA-PD was measured to be approxmately 40 pa when nose was suppressed by averagng suffcently. However, n the absence of averagng, the nose level was ncreased to 30 na; thus, the lowest detectable power P 0.09 W (P av 17 μw). Now let us estmate the power range at the nput port, whch s detectable by the TPA-PD. The loss n the reference branch was estmated to be 0.5 db (L ). On the other hand, the loss n the delay scanner branch was 0.44τ + 1 = 6.3 db (L 2 = 0.23) at the center delay ncludng 0.1 db loss n the wre wavegude, 1 db total loss at two junctons between the LSPCW delay scanner/ dsperson compensator and wre wavegudes and 1.7 db loss at the MMI tap. Therefore, the total loss summng the 2 db couplng loss at the SSC, 3 db total loss between the MMI branch and combner on the n-phase condton, and 0.5 db loss at the juncton between the LSPCW TPA-PD and wre wavegude was 5.5 db. Consderng ths value, the detectable range s P = W. The lowest value wll be lowered by ncreasng the responsvty of the TPA-PD wth a hgher group ndex n addton to the nose reducton by averagng. The hghest value wll be enhanced by further suppressng the nonlnear effects n the delay scanner. Fg. 4. Operaton of two LSPCW components. (a) Delay tunng characterstcs. The nset of the upper panel shows the cross-correlaton waveform of the output pulse measured off-chp, where the FWHM of the reference pulse was 1.0 ps. (b) Responsvty characterstcs of the TPA-PD. Fnally, we operated the entre devce to measure the correlaton waveform of the pulse of 3, 5, 7 and 10 ps n FWHM (blue open and closed crcles n Fg. 5). The ncdent peak power was set at 3.8, 2.3, 1.6, and 1.7 W, respectvely. For comparson, we also calculated the waveform (blue lne) usng Eqs. (3) and (4). Here, we assumed the above L 1 and L 2 and set the cosne functons n Eq. (3) to be unty because we manually measured each plot one-byone, adjustng the delay, phase, and optcal algnment msalgned by the strong heatng. Then the peak to background rato calculated from Eq. (5) becomes 4.9: 1. For the 5- and 7-ps pulses, the measured waveforms were n good agreement wth the calculated ones. For the 3- ps pulse, the plots on the rght sde at lower heatng powers dd not agree wth the calculated waveform. Ths dscrepancy mght stem from the mperfect dsperson compensaton for a wder pulse spectrum. It s understood from the temporal FWHM of the pulse tapped from the delay scanner (red dotted lne). The dsperson was partcularly ncreased on the rght sde. For the 10-ps pulse, the entre waveform could not be acqured because of the lmted delay scannng range. Moreover, the measured correlaton ntensty was unstable due to relatvely low power levels at the tal of the waveform. If we scanned the delay synchronously wth the (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20772

7 phase usng a saw-tooth drve sgnal and measured the waveform repeatedly on an osclloscope, a more stable waveform would be observed after averagng. However, because the heatng power for the delay scannng was as large as hundreds of mllwatts, the detected sgnal at the TPA-PD contaned substantal nose when the heatng power was scanned contnuously and the photocurrent was partcularly weak. Ths nose s caused by the poor heatng effcency to the ar-brdge LSPCW. As a soluton, a slca-cladded LSPCW desgned for tunable delay [17] and a mcro-rng-based tunable delay lne [18], n whch heaters are placed above them, wll be effectve. Fg. 5. Auto-correlaton waveforms (blue) measured usng the developed on-chp optcal correlator (crcles) and a smulated one (sold lnes). Temporal FWHMs of ncdent pulse was 3, 5, 7 and 10 ps from upper to lower pannels. Open and closed crcles depct those obtaned n two dfferent measurement trals. Red dotted lnes show the temporal FWHM of the pulse tapped from the delay scanner and measured externally for each ncdent pulse wdth. 5. Concluson In concluson, we fabrcated an on-chp optcal correlator ncorporatng two photonc crystal slow-lght wavegudes and demonstrated ts operaton. The footprnt of ths devce was only mm 2, whch s substantally smaller than that of conventonal correlators wth freespace optcs. It enables the measurement of optcal pulses shorter than 10 ps; t was lmted to 5 7 ps n ths study but wll be extended to 1 10 ps n future studes by reducng the dsperson and usng wder tunable delay. The detectable ncdent peak power ranged from W but t wll also be extended partcularly n the lower power level by reducng the nose n the TPA-PD. A workng spectrum of up to 10 nm can be acheved by thermally tunng each component. Ths devce wll also enable one-chppng of more advanced optcal measurement nstruments. Acknowledgment Ths work was supported by the New Energy and Industral Technology Development Organzaton (NEDO). (C) 2015 OSA 10 Aug 2015 Vol. 23, No. 16 DOI: /OE OPTICS EXPRESS 20773