- Email: [email protected]
A simple jitter-free picosecond streak camera
Volume 37, number 3
OPTICS COMMUNICATIONS
1 May 1981
A SIMPLE JITTER-FREE PICOSECOND STREAK CAMERA
Wayne KNOX and Gerard MOUROU University of Rochester,
Laboratory for Laser Energetics, Rochester,
NY 14623,
USA
Received 13 January 198 1
Using a DC-biased room-temperature photoconductive switching element, a picosecond resolution optical streak camera is operated at sweep rates up to 7 ps/mm. Single-shot jitter is 2 ps with complete absence of short and long-term drift characteristic of all other sweep drivers. Thishigh accuracy allows the streak camera to be used in a single-shot averaging mode which leads to a sub-picosecond absolute timing accuracy. When operated in this manner, the streak camera becomes a highly reliable and accurate device for measuring transient events on a time scale of single picoseconds. The simplicity of this method makes it highly attractive for many applications in laser fusion and picosecond chemistry and photophysics studies where reliable nodrift operation is highly desirable.
1. Introduction
The picosecond streak camera [ 1,2] is becoming a basic tool in the study of optical transients in the picosecond time scale. Other methods for accessing picosecond time resolution are gated frequency mixing [3], optical Kerr shuttering [4], two-photon fluorescence [5] , and transient absorption spectroscopy [6], but in many cases, use of a streak camera can lead to higher optical sensitivity, time resolution, signal-to-noise ratio, and greater ease in data taking. With the development of streak tubes with resolution of several pitioseconds, the requirements for fast high voltage switching quickly increased. We have demonstrated [7,8] for the first time that high voltage photoconductive switching can exceed all of these requirements and hence provide an excellent streak camera driver. This paper reports on a recent further development in this area which greatly simplifies the use of fast high voltage photoconductive switching applied to the streak camera and results in a subpicosecond timing accuracy and unprecedented stability. In order to achieve a time resolution of a few ps, a photoelectron beam is swept at speeds approaching that of light across a phosphor screen in the streak tube (image-converter tube). This requires an
electrical ramp pulse of several kilovolts (l-5 kV) synchronized with high accuracy to the event to be observed. Conventional electronics can at best offer a jitter of *50 ps for short times with some longterm delaydrift. This comparatively small jitter is obtained with avalanche transistor stacks which produce limited voltage ramps, leading to reduced sweep rates. The highest sweep rates previously obtained were achieved with optically-triggered spark gaps switching up to 16 kV, but with jitter of several hundred picoseconds. At a large streak rate, this jitter is not tolerable. A streak camera was operated with a light-activated silicon switch for the first time with a jitter of +1.7 ps [7,8] using a microsecond-length pulsed bias to avoid thermal instability [8,9]. It was pointed out that a source of this jitter was the small fluctuations in switched voltage due to the statistical error in timing of the Q-switched envelope from the oscillator with respect to the bias pulse which was only flat to within a few percent. In order to resolve this difficulty, a photoconductive switch made of golddoped silicon was biased at 2 kV DC and held at liquid nitrogen temperature in an evacuated vessel to avoid thermal instability [ 10,111. Although this eliminates the need for a pulsed bias, several problems enter with the use of cryogenics. Intrinsic silicon 203
Volume 37, number 3
1 May 1981
OPTICS COfviMUNICATIONS
could not be used because its dielectric breakdown field strength of - 3 X lo5 V/cm at room temperature is dramatically reduced to -lo2 V/cm at liquid nitrogen temperature. This is attributed to a lack of electron-phonon scattering which increases the carrier mean-free-path in the absence of impurities. In order to circumvent this problem, deep impurity centers were added via doping with Au at -1015 ~m--~, thus providing sufficient impurity scattering to maintain an adequate mean-free-path and hence a large dielectric breakdown field strength of about lo5 V/cm. The timing stability of a few ps obtained with the cryogenic Au : Si was dictated by the laser pulse energy and width fluctuations, but the operational advantage of a switching device which was DC biased instead of pulse-biased was clear. The necessity of maintaining the switching device at 77 K complicates its use, and a device which could be DCbiased and operated at room-temperature would be much simpler, compact and portable.
single picoseconds from samples which emit for times long compared to this (ns to ms). In these experiments, the beam must remain deflected for up to ms. Long fluorescence decay times lead to problems in repetitively-scanned systems. Ramps of longer than 1.3 ns can be obtained with a double-stage charging configuration or with the use of a material with a longer recombination time.
2. Experiment The first attempt to switch a large DC voltage using GaAs was made by Lee [ 121, but reduced efficiency was observed and attributed to intervalley scattering. This result was improved upon later with 90% efficiency at up to 8 kV DC [9]. GaAs doped with Cr was chosen because of its high resistivity (-lo8 n-cm) [12] and hence large DC holdoff without thermal instability [9,12] , and its high dielectric strength. A 3 mm gap is biased at O-5 kV DC (fig. 1). When excited by a laser pulse (50-100 llJ at 1.06 m) it becomes conducting with a risetime limited by the laser pulse width [9], and returns to a nonconducting state after about 1.3 ns (l/e). During this time, the deflection plates of an image-converter tube are charged through a current-limiting resistor R, , and remain charged until bias resistor R, returns the voltage to its initial value. With R, = lo* 52 and Cplates = 1O-11 F, this takes about 10-3 s. Thus, the photoelectron beam is swept linearly over 1.3 ns from off-screen on one side to off-screen on the other side, and the beam remains off the screen for a very large time compared to fluorescence decay times of many systems. In many experiments, one is interested in studying emission risetimes on the scale of 204
w ---:fid
II
I -00
I
-30
,Y 0
, 30
I 00
TIME ( P-c) Fig. 1. Experimental configuration. A frequencydoubled Nd3+:YAG single pulse is split and the infrared component (-100 NJ) is directed onto a 3 mm gap of GaAs : Cr, a high resistivity material. When excited by the 30 ps pulse, the GaAs becomes conducting and charges the deflection plates (through Rc) of an imageconverter tube. It returns to a non-conducting state after about 1.3 ns (l/e). The photoelectron beam is thus swept across the screen at speeds of up to half that of light. The deflection plates remain charged until resistor Rg discharges them on the time scale of 1 ms. The sweeping is thus synchronized to the green and W components which provide excitation for samples placed in front of the image converter. The streaked signal is recorded with an intensified OMA and plotted. The sweep speed is easily adjusted by varying the O-5 kV supply or the resistor RC.
Volume 37, number 3
OPTICS COMMUNICATIONS
Since the switching is purely photoconductive, good timing is expected compared to avalanche switching, and in this way, single-shot jitter is reduced to 2 ps at speeds of up to half the speed of light. The sweep speed is easily adjustable from 1 ns full scale to 100 ps full scale by varying the DC bias voltage on the GaAs, and is linear to 5% over 1.3 ns. Signal averaging can be performed with this low jitter to increase further the available timing precision and signal-to-noise ratio. It is found that the timing precision of the system improves with the number of shots averaged. Fig. 2 shows that by using a difference technique [ 111 a time delay as small as 0.40 +-0.05 ps is easily resolvable after averaging as few as 50 shots. This shift of 0.40 ps is the level of timing unstability in the present system after averaging 50 shots. We conclude that the accuracy of this technique is higher than the present system stability. Instability in the acceleration potential in the image converter is a possible source of timing fluctuation, and optical path lengths may need to be stabilized. Laser amplitude and pulse width stabilization will eventually be needed to attain an absolute timing accuracy which is limited by noise. In one run, a time delay of 0.04 ps + 0.02 ps was detected at the limit of S/N = 1 after averaging 100 shots. 2w
1 May 1981
The ability to resolve time shifts of -0.1 ps in luminescence experiments can provide information about relaxation phenomena which are thought to occur on the time scale of lo-l3 s. In addition, the high dynamic range accessed by signal averaging allows detailed studies to be performed virtually in absence of noise characteristic of single-shot data.
3. Conclusion A simple streak camera driver which is DC-biased and operated at room temperature has been devised. Use of the device results in 2 ps single shot jitter, timing accuracy of 0.40 ps when used in averaging mode, and great ease of operation compared to our previous work in pulsed-bias and cryogenic switching. Voltage fluctuations (sweep speed fluctuations) were controlled to less than +O.l%, resulting in exceptional stability and reliability, and the system has no short - or long-term drift. This permits averaging experiments to extend over long periods of time to gain timing accuracy into the femtosecond domain and large S/N ratios for detailed picosecond kinetics studies. It is emphasized that with this technique in many 2w-
3w
3oF-c
15OpJ l.OSP
i I
OPTICAL DELAY LINE
0-5kV DC
uOMA CON TROLLER
TIMING ETALON
&STAGE INTENSIFIER
IMAOE CONVERTER TUBE
Fig. 2. Subpicosecond timing. A time delay of 0.40 f 0.05 ps between two series of 50 shots accumulated under identical conditions caused by instability in the system: (A) the two series of 50 shots, (B) their difference, and (C) their difference magnikd by 10X. The available S/N yields an accuracy of kO.05 ps.
205
Volume 37, number 3
OPTICS COMMUNICATIONS
cases the complex, unreliable and cumbersome electronics commonly used with streak cameras can be replaced with an inexpensive, simple and highly accurate device capable of increasing the usefulness of the streak camera tremendously. It should be mentioned that the performance of this system in terms of timing accuracy scales with the laser pulse width used to excite the Ga-As [8] . With the use of a gigawatt sub-picosecond source currently under construction at LLE, we expect the timing accuracy to be much less than 0.1 ps with appropriate stabilization of accelerating voltages and optical path lengths.
Acknowledgements
This work was partially supported by the following sponsors: Exxon Research and Engineering Company, General Electric Company, Northeast Utilities, New York State Energy Research and Development Authority, The Standard Oil Company of Ohio, The University of Rochester, and Empire State Electric Energy Research Corporation. Such support does not imply endorsement of the content by any of the above parties.
206
1 May 1981
References [ 1] N.H. Schiller, Y. Tsuchiya, E. Inuzuka, Y. Suzuki, K. Kinoshita, K. Kamiya, H. Lida and R.R. Alfano, in: Optical spectra, 14 (1980) p. 55, gives a good introduction to the streak camera. [2] Also D.J. Bradley and G.H.C. New, Proc. IEEE 62 (1974) 313; D.J. Bradley, Opt. Laser Tech. Feb. 1979, Vol. 11, p. 23. [3] B. Kopainsky and W. Kaiser, Optics Comm. 26 (1978) 219. (41 M.A. Duguay and J.W. Hansen, Appl. Phys. Lett. 15 (1969) 192. [S ] D. Von Der Linde, IEEE J. Quant. Electron. QE-8 (1972) 328. [6] R.W. Anderson, D.E. Damshen, G.W. Scott and L.B. Talley, J. Chem. Phys. 71 (1979) 1134. ]71 W. Knox, W.D. Friedman and G. Mourou, post deadline paper I-2, CLEA Washhington, D.C. May, 1979. 181G. Mourou and W. Knox, Appl. Phys. Lett. 36 (1980) 623. PI G. Mourou and W. Knox, Appl. Phys. Lett. 35 (1979) 492. IlO1 N. Stavola, G. Mourou and M. Sceats, Optics Comm. 34 (1980) 409. IllI M. Stavola, G. Mourou and W. Knox, Optics Comm. 34 (1980) 404. I121 C. Lee, Appl. Phys. Lett. 30 (1977) 84.
OPTICS COMMUNICATIONS
1 May 1981
A SIMPLE JITTER-FREE PICOSECOND STREAK CAMERA
Wayne KNOX and Gerard MOUROU University of Rochester,
Laboratory for Laser Energetics, Rochester,
NY 14623,
USA
Received 13 January 198 1
Using a DC-biased room-temperature photoconductive switching element, a picosecond resolution optical streak camera is operated at sweep rates up to 7 ps/mm. Single-shot jitter is 2 ps with complete absence of short and long-term drift characteristic of all other sweep drivers. Thishigh accuracy allows the streak camera to be used in a single-shot averaging mode which leads to a sub-picosecond absolute timing accuracy. When operated in this manner, the streak camera becomes a highly reliable and accurate device for measuring transient events on a time scale of single picoseconds. The simplicity of this method makes it highly attractive for many applications in laser fusion and picosecond chemistry and photophysics studies where reliable nodrift operation is highly desirable.
1. Introduction
The picosecond streak camera [ 1,2] is becoming a basic tool in the study of optical transients in the picosecond time scale. Other methods for accessing picosecond time resolution are gated frequency mixing [3], optical Kerr shuttering [4], two-photon fluorescence [5] , and transient absorption spectroscopy [6], but in many cases, use of a streak camera can lead to higher optical sensitivity, time resolution, signal-to-noise ratio, and greater ease in data taking. With the development of streak tubes with resolution of several pitioseconds, the requirements for fast high voltage switching quickly increased. We have demonstrated [7,8] for the first time that high voltage photoconductive switching can exceed all of these requirements and hence provide an excellent streak camera driver. This paper reports on a recent further development in this area which greatly simplifies the use of fast high voltage photoconductive switching applied to the streak camera and results in a subpicosecond timing accuracy and unprecedented stability. In order to achieve a time resolution of a few ps, a photoelectron beam is swept at speeds approaching that of light across a phosphor screen in the streak tube (image-converter tube). This requires an
electrical ramp pulse of several kilovolts (l-5 kV) synchronized with high accuracy to the event to be observed. Conventional electronics can at best offer a jitter of *50 ps for short times with some longterm delaydrift. This comparatively small jitter is obtained with avalanche transistor stacks which produce limited voltage ramps, leading to reduced sweep rates. The highest sweep rates previously obtained were achieved with optically-triggered spark gaps switching up to 16 kV, but with jitter of several hundred picoseconds. At a large streak rate, this jitter is not tolerable. A streak camera was operated with a light-activated silicon switch for the first time with a jitter of +1.7 ps [7,8] using a microsecond-length pulsed bias to avoid thermal instability [8,9]. It was pointed out that a source of this jitter was the small fluctuations in switched voltage due to the statistical error in timing of the Q-switched envelope from the oscillator with respect to the bias pulse which was only flat to within a few percent. In order to resolve this difficulty, a photoconductive switch made of golddoped silicon was biased at 2 kV DC and held at liquid nitrogen temperature in an evacuated vessel to avoid thermal instability [ 10,111. Although this eliminates the need for a pulsed bias, several problems enter with the use of cryogenics. Intrinsic silicon 203
Volume 37, number 3
1 May 1981
OPTICS COfviMUNICATIONS
could not be used because its dielectric breakdown field strength of - 3 X lo5 V/cm at room temperature is dramatically reduced to -lo2 V/cm at liquid nitrogen temperature. This is attributed to a lack of electron-phonon scattering which increases the carrier mean-free-path in the absence of impurities. In order to circumvent this problem, deep impurity centers were added via doping with Au at -1015 ~m--~, thus providing sufficient impurity scattering to maintain an adequate mean-free-path and hence a large dielectric breakdown field strength of about lo5 V/cm. The timing stability of a few ps obtained with the cryogenic Au : Si was dictated by the laser pulse energy and width fluctuations, but the operational advantage of a switching device which was DC biased instead of pulse-biased was clear. The necessity of maintaining the switching device at 77 K complicates its use, and a device which could be DCbiased and operated at room-temperature would be much simpler, compact and portable.
single picoseconds from samples which emit for times long compared to this (ns to ms). In these experiments, the beam must remain deflected for up to ms. Long fluorescence decay times lead to problems in repetitively-scanned systems. Ramps of longer than 1.3 ns can be obtained with a double-stage charging configuration or with the use of a material with a longer recombination time.
2. Experiment The first attempt to switch a large DC voltage using GaAs was made by Lee [ 121, but reduced efficiency was observed and attributed to intervalley scattering. This result was improved upon later with 90% efficiency at up to 8 kV DC [9]. GaAs doped with Cr was chosen because of its high resistivity (-lo8 n-cm) [12] and hence large DC holdoff without thermal instability [9,12] , and its high dielectric strength. A 3 mm gap is biased at O-5 kV DC (fig. 1). When excited by a laser pulse (50-100 llJ at 1.06 m) it becomes conducting with a risetime limited by the laser pulse width [9], and returns to a nonconducting state after about 1.3 ns (l/e). During this time, the deflection plates of an image-converter tube are charged through a current-limiting resistor R, , and remain charged until bias resistor R, returns the voltage to its initial value. With R, = lo* 52 and Cplates = 1O-11 F, this takes about 10-3 s. Thus, the photoelectron beam is swept linearly over 1.3 ns from off-screen on one side to off-screen on the other side, and the beam remains off the screen for a very large time compared to fluorescence decay times of many systems. In many experiments, one is interested in studying emission risetimes on the scale of 204
w ---:fid
II
I -00
I
-30
,Y 0
, 30
I 00
TIME ( P-c) Fig. 1. Experimental configuration. A frequencydoubled Nd3+:YAG single pulse is split and the infrared component (-100 NJ) is directed onto a 3 mm gap of GaAs : Cr, a high resistivity material. When excited by the 30 ps pulse, the GaAs becomes conducting and charges the deflection plates (through Rc) of an imageconverter tube. It returns to a non-conducting state after about 1.3 ns (l/e). The photoelectron beam is thus swept across the screen at speeds of up to half that of light. The deflection plates remain charged until resistor Rg discharges them on the time scale of 1 ms. The sweeping is thus synchronized to the green and W components which provide excitation for samples placed in front of the image converter. The streaked signal is recorded with an intensified OMA and plotted. The sweep speed is easily adjusted by varying the O-5 kV supply or the resistor RC.
Volume 37, number 3
OPTICS COMMUNICATIONS
Since the switching is purely photoconductive, good timing is expected compared to avalanche switching, and in this way, single-shot jitter is reduced to 2 ps at speeds of up to half the speed of light. The sweep speed is easily adjustable from 1 ns full scale to 100 ps full scale by varying the DC bias voltage on the GaAs, and is linear to 5% over 1.3 ns. Signal averaging can be performed with this low jitter to increase further the available timing precision and signal-to-noise ratio. It is found that the timing precision of the system improves with the number of shots averaged. Fig. 2 shows that by using a difference technique [ 111 a time delay as small as 0.40 +-0.05 ps is easily resolvable after averaging as few as 50 shots. This shift of 0.40 ps is the level of timing unstability in the present system after averaging 50 shots. We conclude that the accuracy of this technique is higher than the present system stability. Instability in the acceleration potential in the image converter is a possible source of timing fluctuation, and optical path lengths may need to be stabilized. Laser amplitude and pulse width stabilization will eventually be needed to attain an absolute timing accuracy which is limited by noise. In one run, a time delay of 0.04 ps + 0.02 ps was detected at the limit of S/N = 1 after averaging 100 shots. 2w
1 May 1981
The ability to resolve time shifts of -0.1 ps in luminescence experiments can provide information about relaxation phenomena which are thought to occur on the time scale of lo-l3 s. In addition, the high dynamic range accessed by signal averaging allows detailed studies to be performed virtually in absence of noise characteristic of single-shot data.
3. Conclusion A simple streak camera driver which is DC-biased and operated at room temperature has been devised. Use of the device results in 2 ps single shot jitter, timing accuracy of 0.40 ps when used in averaging mode, and great ease of operation compared to our previous work in pulsed-bias and cryogenic switching. Voltage fluctuations (sweep speed fluctuations) were controlled to less than +O.l%, resulting in exceptional stability and reliability, and the system has no short - or long-term drift. This permits averaging experiments to extend over long periods of time to gain timing accuracy into the femtosecond domain and large S/N ratios for detailed picosecond kinetics studies. It is emphasized that with this technique in many 2w-
3w
3oF-c
15OpJ l.OSP
i I
OPTICAL DELAY LINE
0-5kV DC
uOMA CON TROLLER
TIMING ETALON
&STAGE INTENSIFIER
IMAOE CONVERTER TUBE
Fig. 2. Subpicosecond timing. A time delay of 0.40 f 0.05 ps between two series of 50 shots accumulated under identical conditions caused by instability in the system: (A) the two series of 50 shots, (B) their difference, and (C) their difference magnikd by 10X. The available S/N yields an accuracy of kO.05 ps.
205
Volume 37, number 3
OPTICS COMMUNICATIONS
cases the complex, unreliable and cumbersome electronics commonly used with streak cameras can be replaced with an inexpensive, simple and highly accurate device capable of increasing the usefulness of the streak camera tremendously. It should be mentioned that the performance of this system in terms of timing accuracy scales with the laser pulse width used to excite the Ga-As [8] . With the use of a gigawatt sub-picosecond source currently under construction at LLE, we expect the timing accuracy to be much less than 0.1 ps with appropriate stabilization of accelerating voltages and optical path lengths.
Acknowledgements
This work was partially supported by the following sponsors: Exxon Research and Engineering Company, General Electric Company, Northeast Utilities, New York State Energy Research and Development Authority, The Standard Oil Company of Ohio, The University of Rochester, and Empire State Electric Energy Research Corporation. Such support does not imply endorsement of the content by any of the above parties.
206
1 May 1981
References [ 1] N.H. Schiller, Y. Tsuchiya, E. Inuzuka, Y. Suzuki, K. Kinoshita, K. Kamiya, H. Lida and R.R. Alfano, in: Optical spectra, 14 (1980) p. 55, gives a good introduction to the streak camera. [2] Also D.J. Bradley and G.H.C. New, Proc. IEEE 62 (1974) 313; D.J. Bradley, Opt. Laser Tech. Feb. 1979, Vol. 11, p. 23. [3] B. Kopainsky and W. Kaiser, Optics Comm. 26 (1978) 219. (41 M.A. Duguay and J.W. Hansen, Appl. Phys. Lett. 15 (1969) 192. [S ] D. Von Der Linde, IEEE J. Quant. Electron. QE-8 (1972) 328. [6] R.W. Anderson, D.E. Damshen, G.W. Scott and L.B. Talley, J. Chem. Phys. 71 (1979) 1134. ]71 W. Knox, W.D. Friedman and G. Mourou, post deadline paper I-2, CLEA Washhington, D.C. May, 1979. 181G. Mourou and W. Knox, Appl. Phys. Lett. 36 (1980) 623. PI G. Mourou and W. Knox, Appl. Phys. Lett. 35 (1979) 492. IlO1 N. Stavola, G. Mourou and M. Sceats, Optics Comm. 34 (1980) 409. IllI M. Stavola, G. Mourou and W. Knox, Optics Comm. 34 (1980) 404. I121 C. Lee, Appl. Phys. Lett. 30 (1977) 84.