Ultrahigh-efficiency second-harmonic generation in nanophotonic PPLN waveguides
Cheng Wang, Carsten Langrock, Alireza Marandi, Marc Jankowski, Mian Zhang, Boris Desiatov, Martin M. Fejer, Marko Loncar
UUltrahigh-efficiency second-harmonic generation in nanophotonic PPLN waveguides
Cheng Wang,
1, 2, ∗ Carsten Langrock, ∗ Alireza Marandi,
3, 4
Marc Jankowski,
3, 5
Mian Zhang, Boris Desiatov, Martin M. Fejer, † and Marko Loncar ‡ John A. Paulson School of Engineering and Applied Sciences,Harvard University, Cambridge, Massachusetts 02138, USA Department of Electronic Engineering, City University of Hong Kong, Kowloon, Hong Kong, China E. L. Ginzton Laboratory, Stanford University, Stanford, CA 94305, USA Department of Electrical Engineering, California Institute of Technology, Pasadena, CA 91125, USA Department of Electrical Engineering and Computer Science,Howard University, Washington DC 20059, USA (Dated: October 23, 2018)Periodically poled lithium niobate (PPLN) waveguide is a powerful platform for efficient wave-length conversion. Conventional PPLN converters however typically require long device lengths andhigh pump powers due to the limited nonlinear interaction strength. Here we use a nanostructuredPPLN waveguides to demonstrate an ultrahigh normalized efficiency of 2600%/W-cm for second-harmonic generation of 1.5- µ m radiation, more than 20 times higher than that in state-of-the-artdiffused waveguides. This is achieved by a combination of sub-wavelength optical confinement andhigh-fidelity periodic poling at a first-order poling period of 4 µ m. Our highly integrated PPLNwaveguides are promising for future chip-scale integration of classical and quantum photonic sys-tems. The second-order nonlinearity ( χ (2) ) is responsible formany important processes in modern optics includingsecond-harmonic generation (SHG) and sum-/difference-frequency generation (SFG/DFG) [1]. Efficient and com-pact χ (2) wavelength converters are crucial elements fora range of applications including entangled photon-pairgeneration [2], quantum frequency conversion [3], low-threshold optical parametric oscillators [4] and supercon-tinuum generation [5]. These devices are typically re-alized using periodically poled lithium niobate (PPLN)crystals, where the periodic domain inversion allows for aquasi-phase-matched (QPM) wavelength conversion pro-cess [6], and lithium niobate (LiNbO , LN) provides large χ (2) nonlinear coefficients (e.g. d = 25 pm/V) [7].Since the nonlinear interaction strength is proportionalto the light intensity inside the device, using waveg-uides with tight optical confinement can dramatically in-crease the conversion efficiencies. Conventional, discrete-component PPLN waveguides however are often based onreverse-proton exchange (RPE) or similar waveguidingtechnologies which only provide small core-to-claddingindex contrast (∆ n ∼ ) [22]. Using standard semiconductormicrofabrication methods, optical waveguides with sub-wavelength modal confinement and propagation lossesas low as 0.03 dB/cm at telecommunication wavelengthshave been realized [23]. The significantly increased opti-cal confinement in principle allows for normalized conver-sion efficiencies to exceed those of RPE PPLN waveguides( ∼ for SHG at 1550 nm) by more than anorder of magnitude. However, the conversion efficienciesof thin-film LN waveguide devices to date have not beenable to provide the promised enhancement, possibly dueto a non-ideal overlap between optical modes and/or im-perfect periodic poling [10–12, 14–16]. While resonatorconfigurations could be implemented to achieve higherconversion efficiencies, they also lead to compromised op-erating bandwidths and less tolerance to environmentalfluctuations [16–20].In this paper, we demonstrate thin-film PPLN waveg-uides with ultra-high normalized conversion efficiencies of2600%/W-cm , more than an order of magnitude higherthan the previous record in PPLN waveguides whilemaintaining a large bandwidth [11]. This is achieved bysimultaneously obtaining sub-wavelength optical confine-ment as well as high-fidelity and repeatable periodic pol-ing. Using a 4-mm-long thin-film LN waveguide, we showcontinuous-wave SHG from telecom to near-visible wave-lengths with a total conversion efficiency of 53% at anon-chip pump power of 220 mW. We achieve the strongnonlinear interaction using first harmonic (FH) and sec-ond harmonic (SH) modes that possess high optical con-finement and large overlap at the same time. Our x-cutLN ridge waveguides have a device-layer thickness of 600nm and a waveguide top width of ∼ a r X i v : . [ phy s i c s . a pp - ph ] S e p z x AirLNSiO (c)(a)(b) P o li ng P e r i od ( µ m ) C a l c u l a t ed η ( % W - c m - ) FIG. 1. (a,b)
Mode profiles ( E z component) for funda-mental TE modes at ∼ ∼
775 nm (b). (c)
Numerically calculated poling period for quasi-phase match-ing (black) and theoretical conversion efficiency (blue) for atypical thin-film PPLN waveguide with a top width of 1440nm. of fundamental transverse-electric (TE ) modes at bothFH ( ∼ ∼
775 nm) wavelengths, whichare used in the conversion process. The TE-polarizedmodes utilize the highest second-order nonlinear tensorcomponent d ( d zzz ) in these x-cut films. In order for thewavelength-conversion process to accumulate over a longdistance, the well-known phase-matching condition needsto be satisfied, i.e. the momenta of the photons involvedin the three-wave mixing process have to be conserved[1]. In the case of SHG, this implies that the phase mis-match ∆ k = 2 k k = 0, where k and k are the wavevec-tors at FH frequency ω and SH frequency 2 ω , respec-tively. In many practical settings, direct phase matchingis challenging and limited to certain wavelengths, polar-izations and mode combinations. In these circumstances,the QPM method is often used, where the domain ori-entation of ferroelectric materials like LN is periodicallyreversed with a period of Λ [9]. This creates an effectivewavevector k QPM = 2 π/ Λ, which is used to compensatefor the phase mismatch k, and to allow monotonic energytransfer over a long interaction length [1]. In the case ofour thin-film LN waveguides, the required poling periodfor first-order QPM is ∼ µ m, significantly smaller thanthat in RPE PPLN waveguides due to a much strongergeometric dispersion [Fig. 1(c)]. While an odd multi-ple of this period could be used for higher-order QPM,it significantly reduces the nonlinear interaction strength[14]. In this work we focus on first-order periodic poling,which requires precise control over the poling uniformityand yield on a micron scale throughout the chip.The numerically calculated SHG efficiencies of ournanophotonic PPLN waveguides exceed 4000%/W-cm [Fig. 1(c)], more than 40 times higher than those in RPEwaveguides [8, 9, 24–26]. This is due to a strong modeconfinement to effective areas of < µ m and a largemodal overlap between the FH and SH modes [Fig. 1(a-b)]. The normalized conversion efficiency η is defined as P out / ( P L ), where P in and P out correspond to FH andSH powers, and L is the device length. We numerically Voltage source w L zx FIG. 2. Schematic of the periodic poling process. Pol-ing finger electrodes are patterned on the surface of an x-cutLN-on-insulator substrate. A high-voltage source is used toperiodically reverse the domain orientation of the thin LNfilm. Ridge waveguides are then fabricated inside the poledregion. Inset shows a false-color SEM image of the fabricatedwaveguide, revealing a poling period of 4.1 µ m with a dutycycle of ∼ calculate η using the following equation: η = 2 ω d n (cid:15) n · A ω A ω (1)where n , n are the effective refractive indices of thewaveguide modes at FH and SH wavelengths, respec-tively, (cid:15) is the vacuum permittivity, c is the speed oflight in vacuum. A ω , A ω are the mode areas at the twowavelengths, defined as: A = (cid:90) Re[ E x H ∗ z − E z H ∗ x ] dxdz (2)where E x,z and H x,z are the electric and magnetic fieldsin the corresponding directions, normalized such that thepeak electric field is 1. To calculate the effective nonlinearcoefficient deff used in Eq. 1, we take into considerationthe full nonlinear susceptibility tensor using the followingexpression: d eff = 2 πA ω (cid:90) (cid:88) i,j,k d ijk E ∗ i, ω E j,ω E k,ω dxdz (3)where i, j, k ∈ x, y, z. According to the symmetry groupof LN (class 3m), the majority of these tensor com-ponents are zero. The remaining non-vanishing coeffi-cients include: d zzz = d = 25 pm/V, d xzx = d xxz = d yyz = d yzy = d zxx = d zyy = d = 4 . d yyy = d yxx = d xxy = d xyx = d = 2 . A ω and A ω ) dramaticallyincrease the conversion efficiency η .We demonstrate high-quality first-order periodic pol-ing in thin LN films to enable efficient QPM nonlinearconversion processes (Fig. 2). The devices are fabri-cated on an x-cut magnesium-oxide- (MgO-) doped LN-on-insulator substrate (NANOLN). We first pattern the SHGLight 10 µm w = 1380 nm w = 1440 nm η ( % / W - c m ) TelecomTLS FPC PPLN OSAEDFA (a)(b)(c) η ( % / W - c m ) Theory - idealTheory - corrected
FIG. 3. (a)
Schematic of the characterization setup. (b)
Measured SHG conversion efficiency versus pump wavelengthsfor two waveguides with the same poling period but differenttop widths. A narrower waveguide gives a longer QPM wave-length, consistent with numerical simulation. (c)
Zoom-inview of the SHG spectral response of the 1440-nm-wide de-vice (solid curve), together with the theoretically predictedresponses (dashed/dotted curves). The green dotted curveassumes a device with ideal poling and structural uniformity,whereas the blue dashed curve shows a corrected transferfunction, which takes into account the actual device inho-mogeneity. Inset shows a CCD camera image of the scatteredSHG light at the output waveguide facet. TLS, tunable lasersource; EDFA, erbium doped fiber amplifier; FPC, fiber po-larization controller; OSA, optical spectrum analyzer. poling finger electrodes using a standard photolithogra-phy and liftoff process. The metal electrodes consist ofa 15-nm Cr adhesion layer and a 150-nm Au layer, de-posited by electron-beam evaporation. We perform theperiodic domain inversion by applying several 5-ms-longhigh-voltage pulses at room temperature. The poled re-gion has a width w = 75 µ m and a length L = 4 mm(Fig. 2). Each poled region can accommodate multipleridge waveguides (3 in our case) without cross-talk dueto the strong optical confinement, allowing for dense de-vice integration. After periodic poling, we remove theelectrodes using metal etchant. We use aligned electron-beam lithography (EBL) to create waveguide patterns in-side the poled regions. The patterns are then transferredto the LN device layer using an optimized Ar + -based dryetching process to form ridge waveguides with smoothsidewalls and low propagation losses [23]. The inset of Fig. 2 shows a false-color scanning electron microscope(SEM) image of the fabricated LN ridge waveguides inthe periodically poled region. Here a hydrofluoric acid(HF) wet etching process is used after waveguide forma-tion to examine the periodic poling quality. At a shortpoling period of 4.1 µ m, the devices still exhibit highpoling fidelity with a duty cycle of ∼ using an end-fire coupling setup shownin Fig. 3(a). Pump light from a continuous-wave tele-com tunable laser source (Santec TSL-510, 1480 1680nm) is coupled into the waveguides using a lensed fiber.An in-line fiber polarization controller is used to ensureTE polarization at the input. When the pump laser istuned to the QPM wavelength, strong scattered SH lightcan be observed at the waveguide output facet using aCCD camera from top of the chip, as is shown in theinset of Fig. 3(c). After passing through the waveguide,the generated SH light is collected using a second lensedfiber and sent to a visible photodetector (EO Systems) forfurther analysis. Figure 3(b) shows the measured SHGresponses of two waveguides with the same poling pe-riod and slightly different top widths of 1440 nm and1380 nm. Here the fiber-to-chip coupling losses of ∼ and 2300%/W-cm , respectively, over anorder of magnitude higher than the best values reportedin previous PPLN waveguides [9–15, 24–26].We show that the measured SHG spectral responseand conversion efficiency can be well explained using acorrected transfer function model. Figure 3(c) shows azoom-in view of the SHG response of the 1440-nm-widedevice. The green dotted curve corresponds to the theo-retically calculated transfer function in the ideal case,showing a sinc-function profile with a maximum nor-malized conversion efficiency of 4500%/W-cm . Com-pared with the ideal case, the measured SHG responseshows a slightly broadened transfer function with a low-ered maximum efficiency, likely due to inhomogeneity ofthe thin-film thickness throughout the 4-mm-long waveg-uide. While a full characterization of device inhomogene-ity would require a phase-sensitive measurement [27], wecan verify that such inhomogeneity is responsible for thediscrepancy between theory and experiment. Since inthe absence of pump depletion, inhomogeneous broad-ening conserves the area of the transfer function, wecompare the areas under the measured and calculatedtransfer functions over the laser tuning bandwidth, whichyields a correction factor of 1.28. Using this number, we C on v e r s i on e ff i c i en cy
53% ConversionDepletion theoryMeasured dataLinear theory 0.1 10.1110 2600%/W-cm Input (mW) O u t pu t po w e r ( µ W ) FIG. 4. SHG total conversion efficiency as a function of inputpower in the pump depletion region, showing a highest mea-sured conversion efficiency of 53% at an on-chip pump powerof 220 mW. Inset shows the input-output power relation inthe low-conversion limit. obtain a corrected transfer function, shown as the bluedashed curve in Fig. 3(c). The corrected transfer func-tion also takes into account the effect from the actualpoling duty cycle of ∼ d eff by 7%. After these corrections, the calculated transferfunction shows good agreement with the measured curvein terms of maximum efficiency, QPM bandwidth, as wellas side lobe position and shape. The remaining discrep-ancy between theory and experiment could be attributedto waveguide losses and unoptimized input polarization.Furthermore, we show that our nanostructuredwaveguides could overcome the traditional efficiency-bandwidth trade-off in bulk LN devices thanks to areduced group-velocity mismatch (GVM). The strongagreement between the measured and theoretically cal-culated transfer functions [Fig. 3(c)] suggests that theQPM bandwidth in our devices is dominated by waveg-uide dispersion rather than device inhomogeneity. In thiscase, the QPM bandwidth is given by ∆ λ = λ | ∆ β (cid:48) | cL ,where ∆ β (cid:48) = ( v − g, ω − v − g,ω ) − is the GVM between theFH and SH optical modes. From the experimental re-sults we can extract a GVM of 150 fs/mm, consistentwith our numerical simulation prediction. The GVMvalue in our thin-film waveguides is half of that in bulkLN ( ∼
300 fs/mm), resulting in a doubled QPM band-width compared with RPE PPLN devices with the samelength. Considering the much higher normalized conver-sion efficiency, our devices show a 50-fold improvement inthe efficiency-bandwidth product over RPE waveguides. Further engineering the dispersion property of these thin-film LN waveguides [28] could lead to devices with evenbroader QPM bandwidths.Finally, we observe the onset of pump depletion andSHG saturation at moderate input powers ( ∼
100 mW) inthe 4-mm device (Fig. 4), verifying our extracted conver-sion efficiencies and demonstrating the high-power han-dling capability of our platform. We use an erbium-dopedfiber amplifier (EDFA, Amonics) to further increase theoptical power from the pump laser. At the output end,we use an optical spectrum analyzer (OSA, Yokogawa)to simultaneously monitor the optical power at FH andSH wavelengths to measure the pump depletion ratio.The highest measured absolute conversion efficiency inour devices is 53%, corresponding to the generation of ∼
117 mW at 775 nm in the waveguide using a pump powerof 220 mW (Fig. 4). At a high SH optical intensity of ∼
10 MW/cm inside the waveguide, we do not observephotorefractive damage of the device after many hoursof optical pumping. The measured pump-depletion be-havior matches well with the theoretical prediction basedon the measured normalized conversion efficiency in thelow-conversion limit. The measured input-output powerrelation in the low-conversion limit follows a quadraticdependence, dictated by the nature of the second-ordernonlinear process (inset of Fig. 4).In conclusion, we show that a nanophotonic PPLN de-vice with sub-wavelength optical confinement and high-quality periodic poling can yield normalized conversionefficiencies that are more than an order of magnitudehigher than in traditional devices. Further increasingthe device length (e.g. 2 - 3 cm) could allow near-unity overall conversion efficiencies with pump powers inthe few-mW range, while maintaining a relatively broadQPM bandwidth. Moreover, our nanophotonic PPLNdevices, together with low-loss optical waveguides [23],tight bends, and high-speed electro-optic interface [29]compatible in the same platform, could provide compact,multi-function and highly efficient solutions at low costfor future classical and quantum photonic systems.Device fabrication was performed in part at the Har-vard University Center for Nanoscale Systems (CNS), amember of the National Nanotechnology Coordinated In-frastructure Network (NNCI), which is supported by theNational Science Foundation under NSF ECCS awardno. 1541959. We acknowledge funding from National Sci-ence Foundation (NSF) (ECCS-1609549, ECCS-1609688,EFMA-1741651); ARL Center for Distributed QuantumInformation (W911NF-15-2-0067, W911NF-15-2-0060);Harvard University Office of Technology Development(PSE Accelerator Award); City University of Hong KongStart-up Fund. ∗ These authors contributed equally to this work † [email protected] ‡ [email protected] [1] R. W. Boyd, Nonlinear Optics (Elsevier, 2008) google-Books-ID: uoRUi1Yb7ooC.[2] D. Bouwmeester, J.-W. Pan, K. Mattle, M. Eibl, H. We- infurter, and A. Zeilinger, Nature , 575 (1997).[3] S. Zaske, A. Lenhard, C. A. Keler, J. Kettler, C. Hepp,C. Arend, R. Albrecht, W.-M. Schulz, M. Jetter, P. Mich-ler, and C. Becher, Physical Review Letters , 147404(2012).[4] N. Leindecker, A. Marandi, R. L. Byer, and K. L.Vodopyanov, Optics Express , 6296 (2011).[5] C. R. Phillips, C. Langrock, J. S. Pelc, M. M. Fejer,I. Hartl, and M. E. Fermann, Optics Express , 18754(2011).[6] C. Langrock, S. Kumar, J. E. McGeehan, A. E. Willner,and M. M. Fejer, Journal of Lightwave Technology ,2579 (2006).[7] I. Shoji, T. Kondo, A. Kitamoto, M. Shirane, and R. Ito,JOSA B , 2268 (1997).[8] Y. N. Korkishko, V. A. Fedorov, T. M. Morozova, F. Cac-cavale, F. Gonella, and F. Segato, JOSA A , 1838(1998).[9] E. J. Lim, M. M. Fejer, and R. L. Byer, ElectronicsLetters , 174 (1989).[10] R. Geiss, S. Saravi, A. Sergeyev, S. Diziain, F. Setzp-fandt, F. Schrempel, R. Grange, E.-B. Kley, A. Tnner-mann, and T. Pertsch, Optics Letters , 2715 (2015).[11] L. Chang, Y. Li, N. Volet, L. Wang, J. Peters, and J. E.Bowers, Optica , 531 (2016).[12] A. Rao, M. Malinowski, A. Honardoost, J. R. Talukder,P. Rabiei, P. Delfyett, and S. Fathpour, Optics Express , 29941 (2016).[13] C. Wang, X. Xiong, N. Andrade, V. Venkataraman, X.-F. Ren, G.-C. Guo, and M. Lonar, Optics Express ,6963 (2017).[14] G. Li, Y. Chen, H. Jiang, and X. Chen, Optics Letters , 939 (2017).[15] R. Luo, Y. He, H. Liang, M. Li, and Q. Lin, arXiv:1804.03621 [physics] (2018), arXiv: 1804.03621.[16] C. Wang, M. J. Burek, Z. Lin, H. A. Atikian,V. Venkataraman, I.-C. Huang, P. Stark, and M. Lonar,Optics Express , 30924 (2014).[17] J. Lin, Y. Xu, Z. Fang, M. Wang, N. Wang, L. Qiao,W. Fang, and Y. Cheng, Science China Physics, Me-chanics & Astronomy , 114209 (2015).[18] R. Luo, H. Jiang, S. Rogers, H. Liang, Y. He, and Q. Lin,Optics Express , 24531 (2017).[19] R. Wolf, I. Breunig, H. Zappe, and K. Buse, OpticsExpress , 29927 (2017).[20] R. Wu, J. Zhang, N. Yao, W. Fang, L. Qiao, Z. Chai,J. Lin, and Y. Cheng, arXiv:1806.00099 [physics] (2018),arXiv: 1806.00099.[21] C. Wang, Z. Li, M.-H. Kim, X. Xiong, X.-F. Ren, G.-C.Guo, N. Yu, and M. Lonar, Nature Communications ,2098 (2017).[22] G. Poberaj, H. Hu, W. Sohler, and P. Gnter, Laser &Photonics Reviews , 488 (2012).[23] M. Zhang, C. Wang, R. Cheng, A. Shams-Ansari, andM. Lonar, Optica , 1536 (2017).[24] T. Sugita, K. Mizuuchi, Y. Kitaoka, and K. Yamamoto,Optics Letters , 1590 (1999).[25] K. R. Parameswaran, R. K. Route, J. R. Kurz, R. V.Roussev, M. M. Fejer, and M. Fujimura, Optics Letters , 179 (2002).[26] R. V. Roussev, C. Langrock, J. R. Kurz, and M. M.Fejer, Optics Letters , 1518 (2004).[27] D. Chang, C. Langrock, Y.-W. Lin, C. R. Phillips, C. V.Bennett, and M. M. Fejer, Optics Letters , 5106(2014).[28] Y. He, H. Liang, R. Luo, M. Li, and Q. Lin, OpticsExpress , 16315 (2018).[29] C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Lonar,Optics Express26