A camera-based three-dimensional (3D) imaging system with a superb time-of-flight (TOF) resolution and multi-hit capability was recently developed for electron/ion imaging [Lee et al. J. Chem. Phys. 141, 221101 (2014)]. In this work, we report further improvement of the event rate of the system by adopting an event-driven camera, Tpx3Cam, for detecting the 2D positions of electrons, while a high-speed digitizer provides highly accurate (∼30 ps) TOF information for each event at a rate approaching 1 Mhits/sec.

Determination of the complete kinematic information of ions and electrons in an ionization/dissociation event requires the knowledge of the full 3D momentum distribution information of the coincident fragments. Therefore, momentum imaging is one of the most powerful tools used in atomic, molecular, optical (AMO), and chemical physics. Over the past few decades, tremendous efforts and resources including a myriad of photoelectron–photoion coincidence spectrometers have been committed to understanding molecular fragmentation and ionization processes.1–12 

The two most common momentum imaging spectroscopies used in understanding reaction dynamics are velocity map imaging (VMI)13 and reaction microscopy (REMI) or cold-target recoil-ion momentum spectroscopy (COLTRIMS).14 The two common types of detectors employed are the 2D MCP/phosphor imaging detector and the delay-line detector for VMI and COLTRIMS, respectively. The 2D MCP/phosphor detector has outstanding multi-hit capability and high event rates but lacks the time resolution; thus, the 3D momentum can only be reconstructed with mathematical transformations, e.g., inverse Abel transform. On the other hand, delay-line detectors had achieved a very good time resolution, but due to longer dead time (∼5 ns), they had a limited multi-hit capability, especially for detecting electrons with small kinetic energy. A few variations of delay-line detectors have been developed to circumvent this issue, among which were the multi-quadrant delay line anode with independent four sets of processing electronics by Hanold and co-workers15 and the delay-line anode incorporated with a phosphor screen to provide positional information by Ismail and co-workers.16 Recently, Li and co-workers developed a hybrid camera-based 3D imaging system that achieved great multi-hit capability and TOF resolution.17–20 This system uses a CMOS camera to measure the 2D positions of electron/ion hits while using a synchronized digitizer to obtain the TOF through full waveform digitization and peak detection. The achieved TOF resolution was 32 ps with a dead time less than 0.7 ns.

Even though the camera employed in the 3D imaging setup is incredibly fast (1 Kframes/s) compared to conventional charge-coupled device (CCD) cameras, it is still not fast enough to operate at an event rate approaching 1 Mhits/s. The highest event rate achieved so far was 2 Khits/s with a laser running at 10 kHz.21 Ultrafast cameras do exist and can achieve 1 Mframes/s. However, these cameras are prohibitively expensive, and their durations of acquisition are usually very short due to the requirement of enormous amount of data storage. Recently, a new type of camera (event-driven) has been developed for both scientific and commercial usage. Instead of capturing frames that contain a fixed number of pixels in a conventional camera, in an event-driven camera, each pixel works independently and can timestamp each over-threshold event with high timing accuracy. Because the output is a stream of over-threshold events (true events) instead of a frame that could be full of zero-value pixels, the data rate has greatly been reduced. The Tpx3Cam22,23 camera was designed to achieve more than 10 Mpixels/s with a standard 1 Gbs Ethernet connection. It should be noted that even though the Tpx3Cam and other pixelated detectors24–26 have achieved a few nanosecond timing resolution, this resolution is not enough for electron TOF measurements. Here, in this work, we demonstrate that when adopting an event-driven camera as a drop-in replacement for a conventional CMOS camera, the camera-based 3D imaging system can achieve an event rate approaching 1 × 106 electron hits per second (Mhits/s) while maintaining its outstanding TOF resolution and very low deadtime for electron detection.

The experimental setup was similar to that used in previous work27,28 but with modifications [Fig. 1(a)]. Two of the major components (Tpx3Cam and a high-speed digitizer) will be briefly discussed. The main components of the Tpx3Cam are the Timepix3 chip29 bump-bonded to a specialized silicon optical sensor (256 × 256 pixels)30 and the Speedy PIxel Detector Readout (SPIDR) readout system.31 A distinct feature of this camera is its capability of providing information on both the Time-over-Threshold (TOT) and Time-of-Arrival (TOA) on every hit detected by using an MCP/phosphor imaging detector. In this experiment, the TOAs were used to timestamp electron events for the synchronization purpose, while the TOTs gave an estimate of the pixel brightness. To stream waveform data in real-time at 1 Mhits/s, a high-speed digitizer was required. An AlazarTech high-speed digitizer (ATS9373) was used to acquire MCP waveforms arising from each electron hit. The ATS9373 is a 12-bit PCIe3 digitizer capable of a sampling rate of 4 GigaSamples/s while sustaining a transfer rate of 6.8 GigaBytes/s to a host computer. Together with the trigger re-arm time (∼200 ns), these features enable the card to capture more than 1 Mwaveforms/s with a time resolution of 250 ps.

FIG. 1.

(a) Schematic of the experimental setup and (b) a typical TOF trace measured from the digitizer.

FIG. 1.

(a) Schematic of the experimental setup and (b) a typical TOF trace measured from the digitizer.

Close modal

To demonstrate the capability of this setup, we deployed the system to measure photo-induced thermionic emission from graphene using a high repetition rate laser system, similar with previously reported work.28 The employed laser was a mode-locked Ti: sapphire oscillator system (a repetition rate of 80 MHz). The center wavelength was 790 nm, and the pulse duration was ∼35 fs. The laser input power was a few tens of mW. Commercially available chemical vapor deposition (CVD) graphene on the fused silica surface (graphenesquare.com) was used without further modification or treatment. The sample was placed in a high vacuum chamber (∼10−9 Torr) at room temperature and was directly mounted onto the first electrode of the spectrometer. The laser power was varied to yield different event rates (100 Khits/s, 200 Khits/s, and 500 Khits/s) as read by the digitizer. The electrons emerging from graphene were accelerated and momentum-focused toward the MCP/P47 phosphor detector (Photonis APD, 75 mm diameter) by using a four-electrode VMI spectrometer. Upon electron impacts, light flashes were produced on the phosphor screen, indicating the hit-positions. The positions were then captured by using the Tpx3Cam camera, and the TOF was obtained by digitizing electrical signals associated with voltage drop in MCP produced by electron hits. The camera was operated in the free-run mode, but the high-speed digitizer was triggered by MCP signals. The signal from MCP was first combined with the laser signal picked-off from a photodiode [Fig. 1(a)] and then digitized. We note that at the observed event rates, the count rate per laser shot was still far below one, which enabled coincident measurement of the position and the TOF of each event. For multi-hit events, the correlation between the peak height of digitizer events and the brightness of camera events (TOT) can be exploited to associate the TOF and the position for each event produced by the same laser pulse.17 

Because the digitizer and the camera cannot be triggered by the laser pulses directly due to the extremely high laser repetition rate, the positional information read from the camera and the TOF from the digitizer will have to be synchronized to provide 3D information (2D position plus TOF) for each event. This was achieved offline by matching the global timestamps of the digitizer events with the TOAs of Tpx3Cam events, both of which were available from the metadata associated with each event. Note that the TOA is not the TOF of the electron hits but a global timestamp registering the time when a camera event is taking place. The TOA has enough depth to run for several hours during the data acquisition, providing the global timestamps with granularity of 1.6 ns, while the digitizer timestamps can be as accurate as 1 ns.

The electron TOF was obtained using a peak detection algorithm on recorded digitizer traces, one of which is shown in Fig. 1(b). The relative time difference between the peaks of the sharp feature (signal from the photodiode) and those of the broad feature (MCP signal) was used as the TOF. We selected the closest peak directly before the MCP signal as the laser timing. Because of the cable length and light path difference, the absolute TOF will have a constant offset from this value, which requires a calibration step to obtain. Here, because the main purpose is to show the instrumentation, this calibration was not performed. The instrument TOF resolution was estimated by measuring the relative delay between two laser pulses, and the standard deviation was about 20 ps, which was similar to previously reported 18 ps18. This was much better than the time resolution of the digitizer, thanks to oversampling of signals. This also suggests that our current TOF measurement scheme should be able to achieve a similar electron TOF resolution at 32 ps.

The 3D measurement results are shown in Fig. 2. Figure 2(c) shows the 2D (X, Y) image as seen by the camera, whereas Figs. 2(a) and 2(b) show unsynchronized and synchronized (X, t) images, respectively. The properly synchronized time-space image of the 3D electron Newton sphere [Fig. 2(b)] confirms previously observed delayed electron emission from graphene.28 The delayed emission has a tail extending beyond 1 ns after the laser irradiation. This was proposed as a signature of long-lived charge carriers in graphene.28 Such a feature was missing in the unsynchronized image [Fig. 2(a)]. This suggests that the developed scheme for synchronizing the Tpx3Cam camera events and the digitizer events worked nicely and achieved a significantly higher event rate at 100 Khits/s. We note the data acquisition time was 30 sec owing to the high event rate while previously taking an hour to accumulate similar counts. To show that it is possible to go even higher event rates, we increased the power of the laser to reach 200 Khits/s and 500 Khits/s. The data at 200 Khits/s show very similar structures [Figs. 3(a) and 3(b)]. However, the data at 500 Khits/s show a truncated tail in the Xt image and a hole appears in the center of the XY image. These features were not due to real dynamics that arose from increasing the laser power. Instead, they were due to the deadtime of MCP. Because of the very high event rate and the small area that electrons hit on the detector, there was a significant chance that the same microchannel was hit consecutively within 1 ms. This hampers the full re-charging of the channel and, thus, reduces the gain. This was confirmed by much smaller clusters on the camera representing single hits at high event rates (Fig. 4). This issue is universal to any MCP based imaging system. One solution is to expand the electron cloud using a smaller acceleration field for electrons. Because the energy of the photoelectrons from graphene is small, especially those arising from delayed emission, such a measure was not effective in the current setup. For this reason, we have not attempted to increase the event rate further. A new setup with a longer time-of-flight length will help solve this. However, it must be emphasized that this issue is not an inherent shortfall of the 3D imaging system presented here. The Tpx3Cam used here is capable of processing 12 Mpix/s (and up to 80 Mpix/s with 10 Gbs optical readout), while the digitizer can acquire > 5 M waveforms/s. Therefore, there are no technical issues to prevent the developed imaging system from achieving 1 Mhits/s when a proper source/spectrometer is employed.

FIG. 2.

Raw 3D electron images at 100 Khits/s. (a) Xt image (t is the TOF) for randomly assigned positions (camera) and TOF (digitizer) without matching event timestamps. (b) Synchronized Xt image after timestamp matching. (c) XY spatial image obtained from the camera.

FIG. 2.

Raw 3D electron images at 100 Khits/s. (a) Xt image (t is the TOF) for randomly assigned positions (camera) and TOF (digitizer) without matching event timestamps. (b) Synchronized Xt image after timestamp matching. (c) XY spatial image obtained from the camera.

Close modal
FIG. 3.

3D raw images of photoemission of graphene at 200 Khits/s [(a) and (b)] and 500 Khits/s [(c) and (d)]. The colormap scale is the same as in Fig. 2.

FIG. 3.

3D raw images of photoemission of graphene at 200 Khits/s [(a) and (b)] and 500 Khits/s [(c) and (d)]. The colormap scale is the same as in Fig. 2.

Close modal
FIG. 4.

Example of (x, y) pixel data from the Tpx3cam corresponding to a 50 µs TOA time slice. (a) and (b) 100 Khits/s and (c) and (d) 500 Khits/s. (a) and (c) TOT (in ns), while (b) and (d) TOA (in ns). Each electron event is represented as a cluster of non-zero pixels. Note that both the size and brightness of the clusters have been reduced from (a) to (c) due to the gain depletion at a higher event rate.

FIG. 4.

Example of (x, y) pixel data from the Tpx3cam corresponding to a 50 µs TOA time slice. (a) and (b) 100 Khits/s and (c) and (d) 500 Khits/s. (a) and (c) TOT (in ns), while (b) and (d) TOA (in ns). Each electron event is represented as a cluster of non-zero pixels. Note that both the size and brightness of the clusters have been reduced from (a) to (c) due to the gain depletion at a higher event rate.

Close modal

To summarize, we have successfully demonstrated that by using the Tpx3Cam, the camera-based 3D imaging system with great multi-hit capability and time resolution is capable of achieving 1 Mhits/s. It is worth pointing out that commercial event-driven cameras with a timing accuracy of 1 μs, which are currently being developed,32 would allow to achieve similar 1 Mhits/s performance at a potentially lower cost.

This research was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), under Award No. DE-SC0012628. We would also like to thank Mr. Chuan Cheng and Professor Thomas Weinacht for their assistance with the Tpx3cam setup.

1.
M.
Richard-Viard
,
O.
Atabek
,
O.
Dutuit
, and
P. M.
Guyon
,
J. Chem. Phys.
93
,
8881
(
1990
).
2.
J. A.
Davies
,
J. E.
LeClaire
,
R. E.
Continetti
, and
C. C.
Hayden
,
J. Chem. Phys.
111
,
1
(
1999
).
3.
G. K.
Jarvis
,
K.-M.
Weitzel
,
M.
Malow
,
T.
Baer
,
Y.
Song
, and
C. Y.
Ng
,
Rev. Sci. Instrum.
70
,
3892
(
1999
).
4.
P.
Downie
and
I.
Powis
,
Faraday Discuss.
115
,
103
(
2000
).
5.
M.
Takahashi
,
J. P.
Cave
, and
J. H. D.
Eland
,
Rev. Sci. Instrum.
71
,
1337
(
2000
).
6.
R. E.
Continetti
,
Annu. Rev. Phys. Chem.
52
,
165
(
2001
).
7.
M.
Lebech
,
J. C.
Houver
, and
D.
Dowek
,
Rev. Sci. Instrum.
73
,
1866
(
2002
).
8.
A. M.
Rijs
,
M. H. M.
Janssen
,
E. t. H.
Chrysostom
, and
C. C.
Hayden
,
Phys. Rev. Lett.
92
,
123002
(
2004
).
9.
G. A.
Garcia
,
H.
Soldi-Lose
, and
L.
Nahon
,
Rev. Sci. Instrum.
80
,
023102
(
2009
).
10.
X.
Tang
,
X.
Zhou
,
M.
Niu
,
S.
Liu
,
J.
Sun
,
X.
Shan
,
F.
Liu
, and
L.
Sheng
,
Rev. Sci. Instrum.
80
,
113101
(
2009
).
11.
B.
Jochim
 et al,
Phys. Rev. A
83
,
043417
(
2011
).
12.
A.
Zhao
,
P.
Sándor
, and
T.
Weinacht
,
J. Chem. Phys.
147
,
013922
(
2017
).
13.
A. T. J. B.
Eppink
and
D. H.
Parker
,
Rev. Sci. Instrum.
68
,
3477
(
1997
).
14.
J.
Ullrich
,
R.
Moshammer
,
A.
Dorn
,
R.
Dorner
,
L. P. H.
Schmidt
, and
H.
Schniidt-Bocking
,
Rep. Prog. Phys.
66
,
1463
(
2003
).
15.
K. A.
Hanold
,
A. K.
Luong
,
T. G.
Clements
, and
R. E.
Continetti
,
Rev. Sci. Instrum.
70
,
2268
(
1999
).
16.
I. M.
Ismail
,
M.
Barat
,
J.-C.
Brenot
,
J. A.
Fayeton
,
V.
Lepère
, and
Y. J.
Picard
,
Rev. Sci. Instrum.
76
,
043304
(
2005
).
17.
S. K.
Lee
,
F.
Cudry
,
Y. F.
Lin
,
S.
Lingenfelter
,
A. H.
Winney
,
L.
Fan
, and
W.
Li
,
Rev. Sci. Instrum.
85
,
123303
(
2014
).
18.
S. K.
Lee
,
Y. F.
Lin
,
S.
Lingenfelter
,
L.
Fan
,
A. H.
Winney
, and
W.
Li
,
J. Chem. Phys.
141
,
221101
(
2014
).
19.
Y. F.
Lin
,
S. K.
Lee
,
P.
Adhikari
,
T.
Herath
,
S.
Lingenfelter
,
A. H.
Winney
, and
W.
Li
,
Rev. Sci. Instrum.
86
,
096110
(
2015
).
20.
C.
Weeraratna
,
C.
Amarasinghe
,
S. K.
Lee
,
W.
Li
, and
A. G.
Suits
,
J. Chem. Phys.
149
,
084202
(
2018
).
21.
L.
Fan
 et al,
J. Chem. Phys.
147
,
013920
(
2017
).
22.
M.
Fisher-Levine
and
A.
Nomerotski
,
J. Instrum.
11
,
C03016
(
2016
).
23.
A.
Nomerotski
,
Nucl. Instrum. Methods Phys. Res., Sect. A
937
,
26
(
2019
).
24.
J.
Long
,
F. J.
Furch
,
J.
Durá
,
A. S.
Tremsin
,
J.
Vallerga
,
C. P.
Schulz
,
A.
Rouzée
, and
M. J. J.
Vrakking
,
J. Chem. Phys.
147
,
013919
(
2017
).
25.
A.
Zhao
 et al,
Rev. Sci. Instrum.
88
,
113104
(
2017
).
26.
A. T.
Clark
 et al,
J. Phys. Chem. A
116
,
10897
(
2012
).
27.
D. A.
Debrah
,
G. A.
Stewart
,
G.
Basnayake
,
J. W. G.
Tisch
,
S. K.
Lee
, and
W.
Li
,
Opt. Lett.
44
,
3582
(
2019
).
28.
L.
Fan
,
S. K.
Lee
,
P.-Y.
Chen
, and
W.
Li
,
J. Phys. Chem. Lett.
9
,
1485
(
2018
).
29.
T.
Poikela
 et al,
J. Instrum.
9
,
C05013
(
2014
).
30.
A.
Nomerotski
,
I.
Chakaberia
,
M.
Fisher-Levine
,
Z.
Janoska
,
P.
Takacs
, and
T.
Tsang
,
J. Instrum.
12
,
C01017
(
2017
).
31.
B.
van der Heijden
,
J.
Visser
,
M.
van Beuzekom
,
H.
Boterenbrood
,
S.
Kulis
,
B.
Munneke
, and
F.
Schreuder
,
J. Instrum.
12
,
C02040
(
2017
).
32.
See http://www.celepixel.com for more information on an example of commercial event-driven cameras.