S Form Apoov AD-A264 635 NTATION PAGE o#,B-,o 07,.0180 . (cid:127) i I(cid:127) ! &WI"Ww~ l ft" a" m ov .wf% I", "WWf mwQ s ,von". i Wemo "j~ i *ft Wf~foa mg.. 'Iw.'JAo wroeuV.u Nlw at W.. *-"a@ 1M boaq"m#1 0"6"£e f Pmsaft.-' .,' 1.Por~w OOA4t*&I 0we ofitmso " A4W%1 i f wa,,f t %on" * *4wlean a) 2. REPORT DATE I. REPORT TYPE AND DATES COVERED Jan 93 Interim Report, I Feb 92-31 Jan 93 4. TITLE AND SUBTITLE 5. FUNDING NUMBERS Subpicosecond D I Electrooptic Sampling i. AUTHOR(S) ii(cid:127) ELECTE 1,. //- David M. Bloom (cid:127) MAY1 4 1993 7,. PFooRMIN ORGANIZATION NAME-, AND ADORESS I. PERFORMING ORGANIZATION Edward L. Ginzton Laboratory REPORT NUMBER W. W. Hansen Laboratories of Physics F -- __ ,- Stanford University Stanford, CA 94305-4028 AFISR.TR. (3'; }58 9. SPONSORING, MONITORING AGENCY NAME(S) AND ADORESS(ES) i0. SPONSORING MONITORING AGENCY REPORT NUMBER Department of the Air Force Air Force Office of Scientific Research (AFSC) r-- "(cid:127)/(cid:127) --F- Bolling Air Force Base, DC 20332-6448 11. SLUPPLEMENTARY NOTES 12a. OtSTRiBUTICN, AVAILABILITY STATEMENT 12b, DISTRIBUTION CODE Unlimited 13. ABSTRACT 'MzimujdQQ wor=) The electro-optic sampling system developed under AFOSR contract no. F49620-85-0016 has been the workhorse for high spepd measurements made in this lab. This report documents the continuing effort to improve this tool and highlights its most recent uses. In addition, we report thp development of a time-lens, a new tool that gives us electronic control of the shape of optical pulses. The development of a time-lers has opened up a new field of research. It is a tool to electronically control the temporal characteristics of optical pulses just like glass lenses control the spatial characteristics. Our results to date include the creation of 7 ps pulses from a CW laser, active focusing (compression) of 55 ps pulses down to 2 ps, and demonstration of the time-reversal properties of time-lenses. A Ph.D. thesis deta'ling the Stanford time-lens qnd nvilse comnress-on exoeriments is included as an appendix. 93-10686 14. SUBjE(cid:127)T TERMS Electro-optic sampling, time-lens, erbium doped fiber ampiltiers 1.3 16. PRICE COD)E "17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITAT,4f) OF AIISTRACT OF REPORT OF THIS PAGE OF ABSTRACT Unclassified Unclassified Unclassified Unlimited Edward L. Ginzton Laboratory W. W. Hansen Laboratories of Physics Stanford University Stanford, California 94305-4028 Interim Report to the Air Force Office of Scientific Research of a Program of Research in Subpicosecond Electrooptic Sampling Contract # F49620-92-1-0099 Principal Investigator: A-X-ci c o -_, -F-0 -,.(cid:127) Accbslo(cid:127)~ Fo, David M. Bloom NTIS CRA&- Associate Professor of Electrical Engineering Vic TA8 .. AvBy hbility Codes January 1993 , dii. S iIDecla Summryx The electro-optic sampling system developed under AFOSR contract no. F49620-85-(X)16 has been the workhorse for high speed measurements made in this lab. This report documents the continuing effort to improve this tool and highlights its most recent uses. In addition, we report the development of a time-lens, a new tool gives us electronic control of the shape of optical pulses. Improvements to electro-optic sampling have been pursued in two areas, new techniques and new pulsed laser sources. In the area of new techniques, the idea of -"slow optics - fast electronics", which uses a CW laser and high speed electrical receiver instead of a pulsed laser, has been explored. A slow opLicb systea consisting of a 1.5 um laser diode and erbium doped fiber amplifier has been constructed and used to make electro-optic tests on GaAs circuits. It appears that the measurement sensitivity using this apnroach is not as good as originally believed, and is inferior to that possible with the original technique. Our work towards an optical synthesizer, a broadly wavelength tunable and precisely controlled laser source has ended because of problems obtaining the fundamental building blocks of our proposed system. In the area of new sources two ideas have been studied. The first is the use of erbium doped fiber lasers to produce short optical pulses, and the second, and more exciting is the development of a "time-lens" to control and focus the temporal shape of optical pulses. The development of a time-lens has opened up a new field of research. It is a tool to electronically control the temporal characteristics of optical pulses just like glass lenses control the spatial characteristics. Our results to date include the creation of 7 ps pulses from a CW laser, active focusing (compression) of 55 ps pulses down to 2 ps, and demonstration of the time- reversal properties of time-lenses. A PhD thesis detailing the Stanford time-lens and pulse compression experiments is included as an appendix. REPORT The electro-optic sampling system as a useful tool The electro-optic sampling system has been used as a key tool in several research projects done by our group. It continues to be the workhorse for the majority of our high speed measurements. Several students are using the system to excite and measure the turn-on characteristics of short gate length MODFETs. Another student used the system to develop a high speed photodetector and microwave detector to build all optoelectronic autocorrelator. This system can replace conventional optical autocorrelators using second harmonic generation to measure short optical pulse widths down to 2 ps. New Techniques for Electro-optic testing In order to build a more commercially practical electro-optic testing system, a new technique was pursued. This technique is a "slow optics - fast electronics -' approach. In the conventional "fast optics - slow electronics" electro-optic sampling system, short optical pulses, and a slow photodetector are used to measure the GaAs circuit waveforms using equivalent time sampling. The drawback to this system is the complex laser needed to generate the short pulses. In the new approach, a simple, commercially available CW laser diode is used to probe the circuit under test. The laser is modulated directly by the circuit waveform, and the modulation is detected by a fast photodiode and electrical receiver. The advantage of this system is that it is less complex, and all components are commercially available. It was seen as a way of making electro- optic testing commercially feasible. To test this idea, we built an electro-optic testing system around the laser diode and fast receiver. A key component to this system is an optical amplifier to boost optical signal level before photodetection. We built an erbium doped fiber amplifier with greater than 30dB of optical gain and good noise performance. The system was tested with an electrical spectrum analyzer as the electrical receiver. This would allow spectrum or frequency domain characterization of the circuit under test. Such a receiver is characterized by a small electrical filter, so that the received noise level is minimized. The system performed as expected, with a bandwidth of 22GHz, limited by the photodector and spectrum analzyer used as the receiver. The more interesting type of system, and the type we have with the traditional approach, is one in which the time domain is measured. To build such a system, we used an electrical receiver which uses a sampler to measure the time domain. The fundamental problem to this type of system, and the one which prevented our system fron working as originally hoped, is the increased receiver noise level in the sampling electrical receiver. The sampling receiver has an equivalent noise bandwidth which is many times larger than the noise bandwidth of the spectrum analyzer. This increased noise level overwhelms the received signal. Without major changes to the commercial sampling receiver, this approach to electro-optic sampling simply will not work. Our work has lead us to the conclusion that the sampling process should be done with optics, rather than with electronics. While both techniques rely on mixing between the sampling pulses and circuit waveforms, there is a distinct advantage to doing this mixing in the optical domain. In the electrical sampling approach, the circuit waveform is imprinted on the CW laser beam. The sampling is done after the light beam has been detected; the photocurrent has an AC component proportional to the circuit voltage. The electrical sampling pulses turn a sampling bridge on and off to downconvert the frequency of the high speed fluctuations to a lower frequency. The key is that the sampling bridge is passive: the diodes in the sampling bridge are either conducting or not conducting. When the diodes are conducting, the RF line is connected to the IF line. The voltage on the IF side of the sampling bridge is never more than that on the RF side, and in practice is lower because of losses. Therefore, the amount of signal power converted to the IF side is proportional to the amount of time that the sampling switch is closed. The longer it is closed, the more power is transferred to the IF port. Since there is always a finite amount of noise in the IF circuitry, the signal to noise ratio at this port is proportional to the amount of time that the switch is closed. To get high time resolution, however, the switch should only be -n for a small amount of time. This is the fundamental limitation of sampling in the electrical domain: time resolution is maximized at the expense of signal to noise ratio. The story is different when sampling in the optical domain. Here, the pulsed laser is modulated by the circuit voltage. The AC component of the detected photocurrent is proportional to the average optical power, and is independent of the pulse duration or repetition rate. Therefore, we can get more signal power by increasing the average optical power, and not suffer a loss in time resolution. We conclude that in electro-optic testing systems, high time resolution measurements should be made by sampling with pulsed laser sources, rather than by sampling electrically. Later, we discuss a new measurement technique which may combine the high time resolution of electro-optic sampling with high spatial resolution. New Sources We pursued two ideas to find new pulsed laser sources. The first was to build a pulsed laser source by modelocking an erbium doped fiber (EDF) laser. These EDF lasers could serve as compact pulsed sources at the 1.5 micron wavelength. We first characterized the pump absorption and gain of a heavily-doped EDF using a Ti-Sapphire laser and then assembled a one- meter-long EDF with a I(X) mW, 980 nm pump diode laser. We achieved a 35% input coupling of pump light into the fiber, resulting in a CW output power of about 2 to 3 mW for a simple linear-cavity laser design. We proceeded to investigate other configurations such as those incorporating a fiber loop reflector. For optimal performance and large bandwidth, we discovered that fusion splices and a ring laser design were needed to minimize reflections. For our first efforts at mode-locking, we designed and developed a bulk-crystal LiNb03 piezo-electric strain- optic (PESO) optical modulator. The PESO modulator could be operated at an acoustic resonance ranging from 1 to about 100 MHz and gave a maximum modulation of 10% for 1.55 micron light. For higher-speed mode-locked operation, we used a commercial 3-GHz integrated Mach- Zender fiber modulator. We generated 30() ps pulses at a 5(X) MHz repetition rate using a 14 m- long, all-fiber ring laser (13.6 MHz fundamental resonance). The total average output power was about 1 mW. The more interesting and fruitful approach to new sources led us into a new field of time- lenses. Time Lens We have developed a tool which gives us electronic control over the temporal shape of optical pulses, much like a glass lens allows us to control the spatial shape of optical beams. This type of control has not been previously available, and should open a new field of research as all the capabilities of such a system are explored. The theoretical basis for this type of system is the fact that the differential equations which describe propagation and diffraction of light beams are similar to those describing the dispersion of light pulses. This similarity has been known since the 1960's and led Kolner to complete the analogy between spatial diffraction and temporal dispersion. This analogy can be extended to include spatial lenses and time lenses. In this case, the temporal analog of a spatial lens ( which provides a spatially varying phase shift ) is a phase modulator ( which provides a time varying phase shift. ) It is relatively easy to build a strong spatial lens by grinding a more rounded glass surface. This results in many radians of phase shift across the face of the lens. It has been very difficult, however, to build phase modulators with much peak phase shift. The key to .ur work was the development of a new optical phase modulator which provides 13 radians of phase shift for 1W of microwave drive power. The details of the space-time analogy and construction of our time lens are in the included Ph.D. thesis by Asif Godil, and are only summarized here. Our time lens is an optical phase modulator operating at 5.2GHz, producing a peak phase modulation of 13 radians for a microwave drive power of IW, and 50 radians for 13W of microwave power. After developing the time lens, we did experiments to confirm the analogy between space and time. We did the same operations on optical pulses that are done on spatial beams: focusing and magnification. In the time domain, focusing a long pulse yields a short pulse, and magnifying a pulse with a single time lens yields a longer, time-reversed pulse. We were able to focus the 55 ps pulses from a modelocked Nd:YAG laser to 6.7 ps with IW of microwave drive power, and 1.9 ps with 13 W of microwave power. These numbers match the calculated temporal resolution of the system. The focused pulse results are shown in Figure 1. The time lens can also be used to create a pulse train from CW light. When used this way, the sinusoidal phase modulation can be thought of as a series of positive and negative time lenses, occurring at twice the microwave drive frequency. Since only the positive lenses can form pulses, the pulse repetition rate is the microwave drive frequency. With this technique, optical pulses can be generated from a CW laser without loosing average power. In addition, the method is stable; since there is electronic control of the light. The results are easily predicted by simple theory, unlike laser modelocking, which still is largely an art form. Our results are shown in Figure 2. The pulses repeat at the 5.2 GHz microwave drive freqi;ency. The measured pulse width is a detector limited 20 ps. The theoretical pulse width is 6.7 ps. When focusing with a spatial lens, decentering the lens with respect to the object being focused should not affect the location of the focused spot. The temporal analogy means that pulse timing jitter should be removed when focusing. This experiment was done by focusing the 55 ps laser pulses to 6.7 ps and measuring the pulse timing jitter before and after focusing. The results are shown in Figure 3. Pulse timing jitter is reduced by and average of 80%. The fundamental limitation is lens aberration, caused by the fact that the phase modulator gives sinusoidal phase vs. time, when quadratic phase vs. time is the ideal. Optical Synthesized Source We have stopped working in this area due to difficulties in obtaining the fundamental technology needed to build such a system. The idea was to use surface emitting lasers (SELs) to build an optical source tunable over 1() GHz with a kilohertz linewidth. SELs are a key component because of their ability to smoothly tune wavelength, unlike distributed feedback lasers which undergo mode hops. We collaborated with Sandia National Laboratories to build the S'ZLs We were not able to build lasers with uniform lasing characteristics such as laser threshold current. We have concluded that SEL technology is not yet mature enough to make the fundamental measurements needed to build the optical sweep synthesizer. Related Work While our work to improve the electro-optic sampling system has been narrowed to the single, exciting topic of time-microscopy, additional work in our lab should be interesting to those needing high speed, high resolution circuit testing methods. This work was supported by ONR/DARPA under contract N00014-92-J-1769. Our new approach combines the high spatial resolution inherent in scanning force microscopy (SFM) with our experience in high speed electrical pulse generation to yield a new ultrafast technique, that is theoretically sensitive enough to map an electric potential profile with better than 1 picosecond time resolution and submicron lateral resolution. It is an non-invasive technique which does not require vacuum. Our efforts are aimed at the challenging task of characterizing high-speed nanometer-scale devices. Presently, no one measurement technique simultaneously addresses the requirements of ultrahigý speed and ultrasmall scale. On-wafer microwave probes and optical sampling systems have limited lateral resolution. Although electron beams can yield high resolution, such systems operate in vacuum and are significantly slower. A scanning force microscope operates by sensing the minute deflections of a cantilever to which is attached an atomically sharp tip. In the non-contact SFM mode of operation, longer range forces (such as an attractive electric force) cause the cantilever to bend. Suppose that a voltage V is present between a conductive SFM tip and the sample device. There will then be an attractive force F experienced by the tip. If we simply model the tip and the device as two plates of a parallel plate capacitor, we obtain: S-"A V2 2 2z where Eo is the permittivity of vacuum, A is the effective area, and z is the effective distance. The key point is that the force is dependent on the square of the voltage. We take advantage of this nonlinearity to extend the measurement bandwidth far beyond the mechanical resonance of the cantilever, unlike past capacitance and potentiometry measurements using an SFM, which were limited by the slow cantilever response and feedback electronics. We avoid this problem by using the square-law force interaction present between the SFM and sample for mixing and sampling. The high-speed signal under test is downconverted to a much lower intermediate frequency. The tip acts as an extremely high-speed mixer/sampler. That is, if two voltages, a signal voltage and an applied sampling voltage, appear between the tip and the sample, there will be a force term corresponding to the product of the two voltages. We have conducted preliminary mixing experiments using a commercial scanning probe microscope (Park Scientific Instruments' Autoprobe) and an available supply of silicon nitride cantilevers which were coated with gold. Our initial results confirm the principles of high- speed operation. We have demonstrated sampling with a time resolution of about 1M ps and mixing for input frequencies up to 20 GHz, above which we were constrained by package and input cable losses. We believe that future high-speed performance will not be restricted by the inherent speed of the nonlinearity or by the extremely small tip-to-sample capacitance but by stray capacitances in the system, which will be critical for measurement frequencies exceeding several hundred gigahertz. In the future, w.- intend to improve time and lateral resolution by integrating high-speed circuitry with a microfabricated cantilever and tip. Publications and Oral Disclosures