Ultrafast spectroscopy has provided researchers with key insights into the mechanisms of chemical processes due to its ability to follow dynamics with femtosecond time resolution. While several techniques have been developed, transient absorption spectroscopy is perhaps one of the most widely applied methods for probing chemical systems. During a transient absorption experiment, a temporally short pump pulse photoexcites a sample and ensuing changes in its absorption spectrum are measured using a time-delayed probe pulse. As photogenerated species typically show distinct absorption spectra, by measuring the time-evolution of different absorption features induced by the pump, information regarding dynamic and chemical processes initiated by light can be tracked as a function of time. Detection of pump-induced changes in the transmission of a probe pulse (Δ𝑇𝑇) is typically achieved by continuously modulating the intensity of the pump using either a mechanical chopper or acousto-optic modulator (AOM). High signal-to-noise in the detection of these changes is then provided by a lock-in amplifier, which isolates the signal components that match the modulation frequency of the pump pulse train.
Over the last decade, high-power ultrafast lasers with pulse generation frequencies in the range of hundreds of kilohertz to megahertz have gained prominence due to their high pulse-topulse stability and faster acquisition rates. [1][2][3] These high repetition rate sources have enabled ultrafast science to push into the area of time-resolved microscopy. [4][5][6][7][8] In a transient absorption microscopy (TAM) experiment, the pump and probe pulses are focused using a high numerical aperture objective to a waist size approaching the diffraction limit. While signals measured in a TAM experiment are typically weaker than those recorded in an ensemble transient absorption measurement due to the lower pulse energies required and smaller volume of molecules being probed, this issue can be offset by the high laser repetition rate, which can allow for the isolation of weak signals through averaging. This averaging is often performed using a lock-in amplifier due to its ability to detect signals at frequencies that match the modulation rate of the pump pulse train. Indeed, TAM studies employing lock-in detection have provided key insights into carrier transport in 2D materials, 5,[9][10][11][12] energy transport in quantum dots [13][14][15] and molecular assemblies, 16- 18 and singlet fission dynamics in molecular crystals. [19][20][21][22][23][24][25] However, while commercial lock-in amplifiers can be readily used for single-channel detection, challenges arise in applications requiring multi-channel data acquisition. This limitation is evident in spectrally or spatially resolved measurements, such as broadband pump-probe spectroscopy and widefield pump-probe microscopy. The frame rate of array detectors and cameras typically employed for multichannel detection cannot match the hundreds of kilohertz to megahertz repetition rates of emerging laser systems used for ultrafast spectroscopy and microscopy, limiting some benefits these high repetition rate systems provide. While schemes for multi-channel lock-in detection have successfully been employed for ultrafast spectroscopy, 26,27 these approaches often present several challenges, such as limitations in the number of acquisition channels resulting in low spectral/spatial resolution, complex electronic triggering schemes, limited repetition frequencies, and/or challenging implementation of detectors. As a result, multichannel lock-in detection schemes are rare within the ultrafast community.
In this report, we describe the design of a transient absorption microscope that performs widefield pump-probe imaging using a camera (Heliotis AG, Helicam C3) wherein each pixel functions as a lock-in amplifier. This camera enables multi-channel, spatially encoded lock-in detection without a complex setup needed by other multi-channel lock-in amplification solutions.
The lock-in camera has previously been employed as a widefield interferometer for portable profilometry, 28 Doppler velocimetry, 29 and non-line-of-sight imaging, 30 and has been used to map modulated changes in the fluorescence of nitrogen vacancies induced by electromagnetic fields. 31,32 Seo and coworkers employed an earlier model of the lock-in camera to perform widefield pump-probe measurements of patterned semiconductor thin films and silicon nanowires. 33 Herein, we expand on their work by both developing a widefield imaging scheme that allows Δ𝑇𝑇/𝑇𝑇 signals to be directly acquired using quadrature detection and demonstrate two widefield imaging modes wherein the focal size of the pump pulse is varied. In the first mode, the pump is tightly focused, which produces a localized excitation whose spatial migration over time can be visualized via a spatially defocused probe. In the second mode, the pump and probe are both defocused, enabling widefield transient imaging to assess how different heterogeneous regions of a sample impact its photoexcited dynamics. As a model system for demonstrating these imaging modes, we employ single crystals of N,N′-bis(2-phenylethyl)-3,4,9,10perylenedicarboximide (EP-PDI), an organic singlet fission material 34,35 that is currently of interest for applications in light harvesting. 36 Our work demonstrates the versatility of lock-in imaging for ultrafast microscopy, which can enable new discoveries in nanoscience, biology, and materials science.
EP-PDI Crystal Growth. Following a similar procedure outlined by Kim and coworkers, 37 self-assembled EP-PDI crystals were grown from a saturated toluene solution heated to 90 °C. Separate 1-mL aliquots of the heated solution were drop cast onto deionized water. As the toluene evaporates, EP-PDI crystals form on the surface of the water solution over the course of twenty minutes. Controlling the temperature of the water bath during crystal formation can be used to control resulting EP-PDI crystal size. Large crystals with a width >10 μm and lengths ranging from 100 -200 μm were grown by drop casting toluene onto a water bath heated to 40 ºC that was allowed to slowly cool to room temperature over the course of one hour. Small crystals with 1-3 μm feature sizes were prepared by drop casting toluene onto water held at ambient temperature. After formation, the crystals were transferred to either a silver mirror or a glass slide for spectroscopic measurements. TAM images of large and small EP-PDI crystals were acquired with a widefield probe following excitation by a tightly focused pump or a widefield pump, respectively, as described below.
Widefield Transient Absorption Microscope. The sample plane of a custom-built transient absorption microscope was imaged by two cameras, a color CMOS camera (Thorlabs, CS165CU) and a lock-in camera (Heliotis AG, Helicam C3), through a 100× objective (Nikon, MPlanAPOHR, 100×/0.90 NA) as shown in Figure 1A. A 90:10 (reflection:transmission) beamsplitter enabled both cameras to image the sample simultaneously with the CMOS and lockin cameras in the transmission and reflection arms, respectively. The CMOS camera, equipped with an infinity focused achromatic lens (f = 200 mm), collected optical microscopy images and aided in aligning the pump, probe, sample, and lock-in camera field of view. The lock-in camera imaged the sample plane through a compound microscope for widefield transient images. Lockin detection was performed at each pixel in a 300 × 300 array. Details of the lock-in detection scheme are described below. An achromatic tube lens (f = 200 mm) created a conjugate imaging plane of the sample in front of a 40× refractive objective (Nikon PlanFluorELWD 40×/0.60 NA) whose aperture was positioned ~2 in. from the detector. That conjugate plane was imaged through the objective onto the detector, yielding a total magnification factor of ~1390× with a conversion ratio of 28.5 nm/pixel and an 8.55 × 8.55 µm field of view. The focus of the optical system was controlled by a piezo objective positioner (Newport, NPO140SG) attached to the 100× objective.
An xy piezo stage (Newport, NPXY200SG or Physik Instrumente, P-517.3CL) and an xy micrometer stage (Newport, 406) controlled the lateral position of the sample.
The transient absorption microscope was deployed in two separate laser systems with high repetition rate lasers. The first laser system was a 1 MHz, 360 fs, 1030 nm ytterbium fiber laser (Spectra-Physics, Spirit 100). An internal AOM was used to downmodulate the repetition rate to a subharmonic of the fundamental rate. The second optical setup employed a 200 kHz, 160 fs, 800 nm Ti:sapphire laser (Coherent, RegA 9000). The repetition rate for each measurement is highlighted in the Results and Discussion section. Other details of each optical setup are provided in the Supporting Information. Briefly, a beamsplitter divided each laser source into separate pump and probe paths. Through a combination of nonlinear optical processes and optical filters, the pump and probe were spectrally tuned to be resonant with the sample absorption profile and target spectral features in the transient signal. The pump and probe were both modulated by AOMs according to the scheme described in the lock-in camera acquisition description below. A delay stage (Newport, DLS325) controlled the pump-probe time delay before both beams were directed into the microscope.
The pump and probe can each be directed into the microscope and imaged by the cameras either in reflection or transmission, as shown in Figure 1A. These configurable modes enable focused and widefield pump transient imaging for both transmissive and reflective samples. Local pump transient imaging (Figure 1B) was achieved by focusing the pump using the 100× objective to the diffraction limit on the sample plane. For widefield pump imaging (Figure 1C) in reflection mode, a lens (f = 1000 mm) focused the pump onto the back aperture of the objective, resulting in even illumination across the field of view at the sample. In transmission mode, an objective (Achromatic 10×/0.30 NA or Nikon PlanFluorELWD 40×/0.60 NA) with a lower numerical aperture (NA) focused the pump onto the imaging plane. The illumination region area could be controlled by moving the low NA objective out of focus with the imaging plane. Widefield illumination of the sample by the probe was achieved in a manner similar to the pump according to the imaging mode. Pump and probe were co-aligned with a 50:50 beam combiner and copropagated in reflection mode and widefield pump transmission mode. A 50:50 beamsplitter routed the beams into the imaging path of the microscope when operating in reflection mode. In transmission mode using a focused pump, the pump and probe counter-propagated through the microscope, with the pump and probe travelling the reflection and transmission paths, respectively.
An optical filter blocked the pump before the lock-in camera tube lens.
Lock-in Camera Acquisition Scheme. Figure 2 shows the image acquisition scheme used for lock-in detection. The lock-in camera collects four frames per reference wave cycle that are differenced to produce an in-phase signal and quadrature signal (Figure 2A). The in-phase signal (blue, frames 1 and 3) is generated by differencing images collected at the peak (frame 1) and minimum (frame 3) of the reference wave. The images acquired and differenced to produce the quadrature signal (yellow, frames 2 and 4) are shifted by a quarter cycle from those used to produce the in-phase signal. The integration period of the lock-in detection is set by the number of reference wave cycles used to generate a lock-in image. Here, the integration period was set to 30 cycles for the data presented below. The reference frequency can be downmodulated to bin several laser pulses into each frame. This approach enables the use of low-energy pump and probe pulses that prevent laser-induced sample damage while maintaining signal intensities on the detector that are large relative to the read noise associated with digitizing images. For the data reported herein, the number of pulses per frame ranged from 15 to 30.
Collection of a TAM image requires measuring two quantities, ΔT, the change in the probe transmission induced by the pump, as well as T, the probe transmission in the absence of the pump.
Dividing ΔT by T normalizes out signal contributions that stem from spatial nonuniformity of the probe. We employ the in-phase difference signal and quadrature difference signal to collect ΔT and T, respectively. In typical lock-in detection, the in-phase and quadrature signals are used to acquire the amplitude and phase of a single incoming signal. To acquire both ΔT and T signals simultaneously with the lock-in camera, the camera is phase synchronized with the AOMs modulating the pump and probe to align ΔT and T to the in-phase and quadrature channels. Instead of considering the in-phase and quadrature channels to be two separate measurements of the same signal, they should be regarded as two separate signals, enabling division to calculate ΔT/T.
Images 1 -4 in Figure 2A highlight which beams are present and which reference wave is used during integration to produce frames 1 -4. To produce the ΔT signal, an AOM modulates the pump such that it is only present during the acquisition of frame 1 (Figure 2A, red) while the probe is present during frames 1 and 3. Differencing these frames produces the ΔT signal in the in-phase image channel. To collect the T signal, a second AOM modulates the probe such that the number of laser shots that contribute to frames 2 and 4 differ, with a smaller number of shots being included in frame 4 (Figure 2A, black). Differencing these frames yields the transmission of the probe, T, in the quadrature image channel. We note that a finite number of shots are collected in frame 4 to prevent saturation of the T image. This saturation is not caused by an individual frame within a cycle, but rather by saturating the accumulated lock-in image acquired by integrating over several frames. The ΔT channel (frames 1 and 3) does not experience this issue as the probe intensity in frame 3 negates probe intensity in frame 1 during integration. However, if the probe is blocked during frame 4, then during integration the T channel can quickly saturate as the probe intensity in frame 2 would not be negated by signal in frame 4. By introducing a limited number of probe shots in frame 4, less than the number of shots in frame 2, the total accumulated intensity in T can be lowered below saturation while still capturing the intensity and profile of the probe. A scaling factor determined by the ratio of shots in frames 2 and 4 can then be used after collection to rescale T to the appropriate intensity. Figure 2B shows simulated images highlighting the ΔT and T lockin images produced by this measurement scheme. The normalized TAM signal, ΔT/T, is calculated by dividing these two images (Figure 2B, right panel). This acquisition scheme maximizes the sensitivity of the transient signal in ΔT and enables acquisition of the normalized transient image ΔT/T while avoiding saturation during lock-in imaging of the probe profile.
The lock-in camera returns unsigned 10-bit format (0 -1023 counts) images with the lockin baseline approximately centered in the middle of the range (500 -600 counts), which varies pixel by pixel. To shift the baseline to zero, a baseline image must be collected for both ΔT and T as they are different. Briefly, the baseline for ΔT is collected with the pump blocked while exposing the probe to the detector while the baseline for T is collected with both beams blocked. Instability in the baseline over time requires a synchronous acquisition of the baseline concomitantly with the transient signal. Details of the baseline acquisition and subtraction methods are outlined in the Supporting Information.
The lock-in camera is capable of a maximum reference wave frequency of 250 kHz (an internal framerate of 1 MHz). The following trigger scheme was used to trigger the lock-in camera and AOMs used for the pump and probe. A digital delay pulse generator (Berkeley Nucleonics, Model 525) was triggered at the repetition rate of the laser. A second pulse generator (Berkeley Nucleonics, Model 525) was triggered by the first at a subharmonic of the laser repetition rate and determined both the reference frequency of the lock-in camera and total number of laser shots per imaging cycle. The lock-in camera anticipates two square pulses per cycle, where the rising edge of each pulse triggers frames 1 and 3 and the falling edge triggers frames 2 and 4. The timing of two internal channels on the pulse generator was set to generate the two pulses per cycle. The internal channels were multiplexed to a single output channel that triggered the lock-in camera. A separate channel was connected to the pump AOM and timed such that the pump was on during frame 1. Another channel triggered the probe AOM such that the probe was on during frames 1 -3. For some of the measurements reported below, the timing was also set to include some probe shots in frame 4 to avoid saturation in T. The repetition rate and reference frequency for each of the measurements below is highlighted at the beginning of its section.
Transient images of an EP-PDI crystal on a silver mirror pumped at 560 nm and probed at 650 nm using the first laser system and the lock-in camera in reflection mode are shown in Figure 3. The repetition rate of the laser was downmodulated by the internal AOM from 1 MHz to 333 kHz. The reference frequency of the lock-in camera was set to 2.78 kHz (11.1 kHz framerate, 30 pulses/frame). Data were averaged for 30 seconds. For these measurements, the pump is focused onto the sample (FWHM, 630 nm) while the probe is defocused (FWHM, 2 μm). With this configuration, the spatially localized excited state population generated by the pump can be contrasted with the surrounding ground state population within a single image. This result is observed in the in-phase image, which shows an excited state absorption feature in the transient signal, ΔT, of the EP-PDI crystal at a pump-probe time delay of 5 ps (Figure 3A). The widefield image of the probe intensity profile, T, collected as the quadrature image is shown in Figure 3B. The probe is large enough to illuminate a portion of the sample including the edge of the PDI crystal, visible in the upper left corner of the image. Diffraction rings within the image arise from out of focus dust in the imaging setup and do not impact the normalized transient signal, ΔT/T (Figure 3C), which is calculated by dividing the in-phase image by the quadrature image. This While noise can arise from sources such as laser instability, pixel readout noise (σpixel) dominates for the lock-in camera. This readout noise can be determined for each pixel by calculating its standard error from a sequence of lock-in images. Noise for individual pixels is also well approximated by calculating the standard deviation of a region of pixels in an averaged transient image at a baseline signal level. Noise fluctuations from the same pixel from the in-phase and quadrature images are not correlated and cannot, therefore, be used for correcting noise between both images. However, σpixel is the same for both lock-in images. The estimated noise for a pixel in the normalized transient absorption image, 𝜎𝜎 Δ𝑇𝑇/𝑇𝑇 , is calculated using Eq. 1:
which, with the condition that Δ𝑇𝑇/𝑇𝑇 ≪ 1, can be approximated with Eq. 2:
Thus, the noise level for each pixel in Δ𝑇𝑇/𝑇𝑇 is primarily dependent on the intensity of the probe in the quadrature image, T, within the dynamic range of the detector. This is evident in Figure 3C,
where regions of low probe intensity around the edges of the transient image shown in Figure 3B manifest higher pixel noise.
The pixel noise level is calculated from a baseline region of the in-phase image shown in
. This transient image is a composite from 30 seconds of averaging and results in a noise level of σpixel = 0.15 counts. With an approximate probe intensity of 200 counts at the pump location, 𝜎𝜎 Δ𝑇𝑇/𝑇𝑇 for each pixel is 7.5 × 10 -4 Δ𝑇𝑇/𝑇𝑇. This level of sensitivity is on par with previous reported widefield transient measurements. 23,38 However, this value falls below that achievable using a single photodiode, standard lock-in amplifier, and an averaging period of a few seconds at a single position, which can yield sensitivities ranging from 10 -5 to 10 -7 Δ𝑇𝑇/𝑇𝑇. 1,39 While the sensitivity level of a single pixel within the lock-in camera is lower, the advantage of collecting a 300 × 300 pixel field-of-view provides a significant reduction in acquisition time. Sensitivity can be increased either by employing a longer averaging period or increasing the probe intensity. Note, when increasing sensitivity by adjusting the probe intensity, care must be taken as undesired signal effects, such as cross-phase modulation or other coherent artifacts, can be caused by setting the intensity of the probe close to that of the pump. The contrast-to-noise ratio (CNR) characterizes the relationship between the signal level and the noise of a system and is calculated as CNR = 𝐴𝐴/𝜎𝜎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 where 𝐴𝐴 is the amplitude of the signal. 40 It is dependent on pump power, sample properties, and time delay. The CNR for the transient image shown in Figure 3C is ~33.
Transient imaging using a focused pump pulse enables observation of the spatial distribution of photoexcited species. By scanning the delay between the pump and probe, timedependent changes in this spatial distribution can be imaged. 5,8,18,23,38,39,41,42 To demonstrate the lock-in camera's ability to track photoexcitations, transient transmission images of a large EP-PDI crystal on a glass slide were produced using focused pump pulses at 450 nm and widefield probe pulses at 600 nm generated by the first laser system (Figure 4). The laser repetition rate was 1
MHz and the reference frequency of the lock-in camera was set to 10 kHz (40 kHz framerate, 25 pulses/frame). Data were averaged for 30 seconds at each time delay. The pump was focused tightly (FWHM, 430 nm) onto the sample using the 100× objective. The probe, counter propagating with respect to the pump through the sample, was loosely focused (FWHM, 9 μm) by the 10× objective and imaged through the microscope onto the lock-in camera. delay times. Each image is normalized to the peak signal at 3 ps with the normalization factor shown in the image. At each time delay, the center portion of the signal is dominated by a positive Δ𝑇𝑇/𝑇𝑇 signal that we attribute to the ground state bleach associated with excitons produced within the EP-PDI crystal. Surrounding this center portion is a weak negative Δ𝑇𝑇/𝑇𝑇 signal that arises from a photo-induced change in the crystal refractive index that alters the path of the probe as it transmits through the sample. This phenomenon has previously been discussed in detail by Ashoka package developed by these authors that can be parameterized to match the experimental conditions of a transient imaging microscope to determine its precision and accuracy. 40 Spatial calibration of the lock-in camera indicates a resolution of 28.5 nm/pixel. As inputs we specify a viewing window of 3.7 μm over 130 pixels with a signal FWHM of 627 nm over 22 pixels. The input CNR was set to 33, as determined above. A decay constant of 100 ps and a diffusion length of 30 nm were set as sample parameters. In total, 80 pump-probe time delays were acquired to match the data set shown in Figure 4. The proportion of simulated data sets whose fit diffusion constant was within 50% of the nominal diffusion constant (30 nm) was 0.69. This value rose to 0.98 for a nominal diffusion constant of 50 nm. These length scales are on par with measured diffusion lengths in organic electron donor and acceptor materials. 44 We note that further improvements in resolution can be gained by increasing the CNR and oversampling in the time domain.
Often microscopic-scale structural and morphological heterogeneity like crystalline grain boundaries and spot defects can have a significant influence on the bulk properties of a material. [45][46][47][48][49] Widefield TAM enables imaging of spatially dependent phenomena such as charge transfer, energy transport, and exciton dynamics at the sub-micron scale, without the need to carry out point-by point scanning, reducing the total acquisition time for a field of view from hours to minutes. We demonstrate widefield transient imaging of a smaller EP-PDI crystal on a glass slide using the lock-in camera and the second laser system (Figure 5). The laser repetition rate was 200 kHz and the reference frequency of the lock-in camera was set to 3.33 kHz (13.33 kHz framerate, 15 pulses/frame). Data were averaged for 30 seconds at each time delay. Rather than employing a tightly focused pump, the pump was defocused to photoexcite the entire region imaged by the probe in order to detect spatial variations in the transient signal across the crystal. For these measurements, a 400-nm pump pulse was used to photoexcite the sample, producing spin-singlet excitons that subsequently undergo singlet fission. This process was monitored using a probe pulse centered at 750 nm that tracked changes in the intensity of an induced absorption band associated with photogenerated singlet excitons. 34,35,50 5B displays an image of the crystal collected by the lock-in camera in normal imaging mode. Both It is evident in the transient images that there is an enhancement of the signal near the edges of the EP-PDI crystal. This enhancement may be caused by a disrupted assembly of the EP-PDI molecules where the structural constraints found in the bulk of the crystal are relaxed, leading to different intermolecular electronic couplings and ultimately different singlet fission dynamics.
Indeed, differences in the rate of singlet fission that occurs at the edges and in the bulk of molecular crystals have been reported. For example, distortions to the crystalline packing of molecules at the edges of TIPS-pentacene crystals have been shown to accelerate their ability to undergo SF. 25 It has also been shown that singlet fission can be enhanced near the edges of rubrene crystals due to the introduction of structural disorder that lowers the symmetry of the crystal. 24 We note that EP-PDI exhibits polymorphism, 51 which suggests unique structures could be formed at EP-PDI crystal edges that may impact their dynamics. This scenario could not only be responsible for the differences in dynamics observed within the bulk and at the edge of the crystal, but may also explain the change in the sign of the signal near the top portion of the crystal. While a full analysis of the intermolecular structure of the EP-PDI crystal edges is beyond the scope of this work, the images in Figure 5 demonstrate the lock-in camera's ability to spatially resolve distinct excited state populations within a heterogeneous material and track their ensuing dynamics.
In this work we have described the use of a lock-in camera for widefield transient