1. Introduction

Mounting interest in high resolution, label-free, non-destructive imaging of biological and manufactured samples has led to a resurgence in the technique of photoacoustic (PA) microscopy (PAM) [1–3]. In typical optical-resolution PAM (OR-PAM), a nanosecond or picosecond pulsed laser is focused to a diffraction limited spot on the sample through a series of lenses or a microscope objective. Absorption of a portion of the incident beam results in a localized isochoric heating and subsequent thermoelastic expansion, leading to the emission of an acoustic wave that is typically detected using an ultrasound (US) transducer [4]. While the potential for simultaneous pulse-echo US imaging in current OR-PAM systems exists, most systems use US transducers with central frequencies in the 20-60 MHz range, resulting in large acoustic focal spots and poor resolution US images [5]. However, when using transducers with central frequencies in the hundreds of megahertz (MHz) range, such as those used in scanning acoustic microscopy, the lateral resolution achievable with US imaging approaches (and in some cases, surpasses) that of OR-PAM [6,7]. Furthermore, since the axial resolution in PAM is proportional to the transducer bandwidth [5,8], the use of ultra-high frequency (UHF) transducers also results in micron level axial resolution [9,10].

With UHF-PAM systems, sequential acoustic and PA imaging is readily achievable [11]. While the PA image derives contrast from the spatial variation in the sample’s optical absorption properties, the contrast in US imaging is dependent upon the sample’s mechanical properties (e.g. acoustic impedance, attenuation, speed of sound) [7]. We have previously demonstrated that the unique information contained in both image types can be used to differentiate between different kinds of leukocytes in a human blood smear [11]. By tuning the illumination wavelength in the PA measurements, unique absorption profiles for stained lymphocytes, neutrophils, and monocytes were obtained. Further, differences in the acoustic attenuation through each cell type suggested variation in the intracellular chromatin structure.

In some OR-PAM setups, triplex imaging can be realized without modification to existing hardware. For example, Subochev et al. reported a triplex reflection-mode PAM setup capable of simultaneous diffuse optical reflectance (DOR), PA, and US imaging [12–14]. In this system, photons from the PA excitation beam backscattered from the sample are detected through absorption in the 35 MHz PVDF transducer, providing a detectable signal dependent on the sample DOR. The same photons absorbed in the hemispherical PVDF result in the emission of a laser induced US wave, which propagates towards the sample and can be used for US imaging. However, due to the low central frequency of the PVDF transducer, the resolution of the PA, US, and DOR images is limited to 50 μm, 35 μm, and 3.5 mm, respectively, and thus the system is not suitable for imaging samples with micron level detail [12]. Furthermore, due to the mechanism of acoustic wave generation, the amplitude of the laser induced US pulse is directly coupled to the DOR signal, and therefore cannot not reliably be used to assess the mechanical properties of the sample.

Here we present a new method for simultaneously generating images that depict the mechanical, optical absorption, and optical attenuation properties of a sample with micron-level detail. Mechanical and absorption properties are assessed via UHF pulse-echo US and PAM, respectively, while the optical attenuation properties are provided by a novel sensing technique termed Photoacoustic Radiometry (PAR). The PAR technique is readily implemented in PAM setups in which the laser used for PA signal generation illuminates the transducer after passing through the sample. Optical attenuation of the laser beam by the sample influences the amplitude of the detected PAR signal, providing image contrast dependent on both optical absorption and scattering. In this work, we describe and characterize the PAR technique, and demonstrate its ability to acquire high-resolution images of optically transparent samples. We then demonstrate duplex micron-resolution PAR/PA imaging of single biological cells, and triplex PAR/PA/US imaging of an integrated circuit die on a silicon wafer.

2. Methods

2.1 Description of the PAR technique

Typically there are two main components in single element UHF transducers: a multi-layer active element, used to generate the US plane wave; and a sapphire or quartz buffer rod, which is used to focus the resultant wave [7]. In these implementations the active element consists of a piezoelectric material (e.g. zinc oxide, ZnO), sandwiched between two thin metal films which are used as electrical contacts [7,15,16]. Application of a potential difference results in an expansion of the piezoelectric layer and the emission of an acoustic wave. Conversely, compression of the piezoelectric layer within the active element due either to backward propagating US waves (Fig. 1(c)), or forward propagating PA waves (Fig. 1(b)), results in a detectable voltage. In the PAR technique (Fig. 1(a)), the portion of the excitation laser beam that is transmitted through the sample impinges directly upon the active element, generating a PA wave that is detected by the piezoelectric material. In this way, for every laser pulse emitted, two PA waves are generated simultaneously; one originating from the sample, and the other originating within the transducer. As ZnO is transparent to visible light [17], we hypothesize that the PAR signal is due to laser absorption in the metallic contacts. The PAR signal is directly proportional to the laser fluence at the active element, and is thus sensitive to both the optical absorption and scattering properties of the sample being interrogated.

 

Fig. 1 a) An illustration of the PAR technique. A laser pulse, having been attenuated by the sample, strikes the active element of the transducer, and produces a PAR signal at time t₀. b) The typical scheme for photoacoustic signal generation. The PA signal is detected at time t₁. c) The US pulse-echo signal is detected at time 2t₁. d) A representative RF line acquired from a triplex scan of a permanent marker on a microscope slide. Various features of the signal, including excitation pulses from the Waveform Generator (WFG) and US pulse generator, are labelled. Multiple detections of reflected acoustic waves trapped within the transducer are denoted with subscripts. e) – g) Time-gated PAR, PA, and US signals, respectively.

Download Full Size | PDF

2.2 System specifications

The UHF-PAM setup consists of a SASAM acoustic microscope (Kibero GmbH, Germany) modified to include a pulsed laser source. A schematic representation of the system is shown in Fig. 2(a). The SASAM operates in transmission mode, with a 10X optical objective with numerical aperture of 0.25 (Olympus, Japan) situated below the sample, and the transducer used for signal detection aligned confocally with the objective on the opposing side. The system was equipped with one of two interchangeable lasers: a 532 nm pulsed Nd:YAG laser (Teem Photonics, France) with a pulse width of 330 ps and pulse repetition frequency (PRF) of 4 kHz, or a 1064 nm laser (Teem Photonics, France) with a pulse width of 500 ps and PRF of 2 kHz. The pulse energies used for image acquisition with the 532 nm and 1064 nm lasers were 30 nJ and 450 nJ, respectively. A 70:30 beam splitter (Thorlabs, USA) was used to sample the incident beam energy and direct it to a joulemeter (Gentec-EO, USA) for pulse energy measurement. Signal detection was performed using a 200 MHz transducer with a 30° semi-aperture angle, −6dB bandwidth of 120 MHz, and depth-of-field of 54 μm. All acquired signals were digitized at a rate of 8 GS/s with a 10-bit digitizer (Agilent, USA). In experiments where only PAR images were generated, 100x signal averaging was used to maximize SNR. All data acquisition was performed at 37°C, and MilliQ water was used to provide acoustic coupling between the transducer and sample when PA or US signals were acquired in addition to the PAR signal. For all experiments, the features which were imaged using the PA and US modalities were on the surface of the sample, in direct contact with the acoustic coupling medium and closest to the US transducer.

 

Fig. 2 a) Experimental system setup, consisting of a PA microscope in transmission mode, equipped with an UHF transducer. b) Amplitude of the recorded PAR signal as a function of incident laser pulse energy for direct transducer illumination. c) Edge spread function (ESF) and the corresponding line spread function (LSF) acquired by scanning the edge of one of the elements on a USAF test target. The FWHM of the LSF is taken as the lateral resolution of the system.

Download Full Size | PDF

2.3 PAR imaging of a transparent sample

A glass bottom petri dish with etched alphanumeric patterning (MatTek, USA) was scanned with the 532 nm laser. The glass was composed of boroscilate and had a thickness of approximately 170 μm. For the first scan, the transducer and dish were separated by an air gap. Immediately after this non-contact scan, a drop of Microscope Immersion Oil (Olympus, USA) with refractive index of 1.516 was added, completely filling the air gap between the etched glass and transducer. A second scan was then performed over the same region of interest. Both raster scans were 600x600 μm in size with a step size of 4 μm in both fast and slow directions. Maximum amplitude projection (MAP) images of the PAR signals were formed for each scan.

2.4 Simultaneous PAR and PA imaging of biological samples

CAKI-2 renal carcinoma cells (ATTC, USA) were cultured in a glass bottom petri dish (MatTek, USA) in McCoy’s 5A medium supplemented with 10% fetal bovine serum. The thickness of the glass on the bottom of the dish was approximately 170 μm. The media was aspirated, and the dish was rinsed with 1X PBS. The cells were then fixed with 2 mL of ice cold methanol for 10 minutes. The methanol was aspirated, and a second PBS rinse was performed. A 1:200 dilution of DRAQ5 dye (BioStatus, UK) in PBS was added to the fixed cells to stain the cell nucleus. The cells were incubated in the dark with the dye solution for 10 minutes. To remove residual dye after the staining process, three PBS rinses were performed. Cells of interest were identified using brightfield microscopy, and PA/PAR measurements were performed using the SASAM and the 532 nm laser. Although DRAQ5 exhibits an absorption maximum at 647 nm, the dye absorbs at wavelengths as short as 488 nm [18], and thus can be excited in PAM with the 532 nm laser. After the scan, the sample was transferred to a Zeiss Axio-observer microscope where DIC and fluorescence images of the cell were acquired with a 40X, 0.6 NA objective, and an OrcaR2 CCD camera. DRAQ5 exhibits low photobleaching [19], and there was no observable decrease in fluorescence intensity for the scanned cell compared to other cells in the dish. The scan region for the biological cell was 80x80 μm. The cell was translated through the overlapping optical and acoustic focal zones in a raster pattern, with a step size of 0.5 μm in the fast and slow directions. Acquired PAR and PA signals were isolated by time gating, and a MAP image was formed for each data set. A binary mask for the PA MAP image was generated by computing the sum of the absolute value of the RF-line for each scan position, and then zeroing pixels with value lower than the interior of the cell nucleus.

2.5 Triplex imaging of an integrated circuit

As a proof of concept for the triplex PAR/PA/US technique, an integrated circuit die on top of a 400 μm thick silicon wafer (TMS320C51 DSP, Texas Instruments, USA) was imaged. As silicon is semi-transparent in the near-IR [20], a 1064 nm laser was used to maximize transmission through the wafer substrate to the integrated circuit on the top surface. A TTL signal generated upon laser emission was used to trigger the US pulser for simultaneous pulse-echo US measurements. Nanosecond jitter in the triggering of the pulser decreases the SNR of the US signal when high averaging is used, thus only 5x signal averaging was used for all triplex signal acquisition. A 500x500 μm region of the die was raster scanned using a step size of 2.5 μm. The resultant scan data set was time gated to isolate and generate MAP images for each of the PAR, PA, and US signals, respectively.

3. Results & discussion

3.1 PAR characterization

A representative triplex RF line acquired from ink on a glass coverslip is shown in Fig. 1(d). The PAR, PA, and US signals are labelled with arrows. In addition to the signals used for image generation, several other features are noted. Excitation pulses from both the waveform generator (WFG) and US pulser are observed after detection of the PAR signal, but before detection of the PA signal. The large impedance mismatch between the buffer rod and coupling medium results in a strong echo of the generated US pulse from the transducer rim. This mismatch also causes multiple reflections of acoustic waves within the transducer. Finally, there are some unidentified artifacts in the RF-line; however, the time at which these artifacts occur is invariant, and thus they do not compromise the signals containing information about the sample. As demonstrated in Figs. 1(e)-1(g), each of the PAR, PA, and US signals has high signal-to-noise ratio (SNR) and can be isolated through time gating. The amplitude of the PAR signal as a function of laser pulse energy was assessed by removing the sample stage and directly illuminating the transducer. The maximum PAR signal amplitude as a function of laser pulse energy is shown in Fig. 2(b). The PAR signal amplitude increased linearly (R² = 0.994) with increasing laser energy, supporting the hypothesis that the PAR signal is due to PA excitation. The SNR of the PAR signal, defined as the maximum PAR signal voltage divided by the root mean square of the noise in the same RF-line, was greater than 14 dB for pulse energies above 1 nJ. In addition to determining the minimum pulse energy required for a usable PAR signal, these measurements also demonstrate that the PAR signal originates completely within the transducer; i.e. that the detected signal is not the result of any PA or US waves which originate from the sample and are detected as multiple reflections (c.f. Figure 4.4 in Briggs [7]).

To determine the resolution of the PAR and PA imaging techniques, the edge of one of the elements on a USAF test target was scanned with a step size of 0.2 μm. The resultant edge spread function (ESF) is shown in Fig. 2(c). For enhanced visualization, only a subset of the scan points acquired are displayed on the graph. The ESF was fit to an error function, and the line spread function (LSF) was calculated via differentiation. The size of the acoustic focal spot for the 200 MHz transducer is approximately 8 μm, thus the 1.65 μm FWHM of the LSF can be taken as the lateral resolution for both PAR and PA techniques, and is close to the diffraction limit for a 0.25 NA objective with an incident wavelength of 532 nm.

3.2 PAR imaging of a transparent sample

A brightfield image of the letters etched on the surface of the empty glass bottom dish is shown in Fig. 3(a), with the corresponding non-contact PAR image shown in Fig. 3(b). The glass is transparent with negligible optical absorption at 532 nm. Thus, no PA signals can be acquired from the sample, and it is impossible to image any of the object features using PAM techniques alone. In both the brightfield and PAR images, image contrast can be attributed to the scattering of light due to variations in geometry and refractive index. The most striking feature in the PAR image is the high contrast at the vertical edges of the etched channels. This can potentially be explained by total internal reflection of a portion of the focused laser beam at the vertical channel wall. The maximum angle of incidence of the beam to a surface orthogonal to the objective’s optical axis (≈14.5°) is well below the critical angle for an air/glass interface (≈41°); however, when the surface is parallel with the optical axis, as in the case of the vertical channel walls, the incident angle is 75.5° and a majority of the incident beam can be reflected back into the glass. Such a reflection would decrease the fluence at the transducer active element and in turn decrease the amplitude of the recorded PAR signal. Debris on the top surface of the glass appeared as dark spots in the PAR image with signal amplitude substantially lower than that of the background. In comparison, debris located on the bottom of the dish away from the objective focal plane (denoted with arrow heads) had a blurry appearance with higher PAR amplitude than those on the top. The PAR image acquired from the same scan region after the addition of immersion oil with index of refraction similar to the glass is shown in Fig. 3(c). The etched letters are no longer visible, as the refractive index mismatch at the channel wall has been eliminated. The only discernable features are the dark spots caused by the debris on the bottom of the dish. After the addition of the immersion oil, the mean signal amplitude in the image was observed to increase approximately by a factor of three. This could potentially be due to decreased refraction at the surface of the sample, and transducer rim, leading to an increase in the fluence at the transducer active element. In the future, we plan to investigate this phenomenon further with the use of Finite difference time domain (FDTD) or Finite Element Method simulations.

 

Fig. 3 a) Brightfield microscopy image of an etched glass bottom petri dish. The letters “P” and “A” are shown. b) Non-contact PAR scan of the same letters shown in a). Excellent contrast is observed at the edges of the etched channels. Two out-of-focus pieces of debris on the bottom of the dish are indicated with arrowheads. c) PAR image of the same region after the addition of microscope immersion oil with a refractive index similar to the glass. The etched letters are no longer visible; however, the out-of-focus pieces of debris can still be seen clearly. The scale bar in c) is 100 μm and is applicable to all images.

Download Full Size | PDF

3.3 Simultaneous PA and PAR imaging of single biological cells

A differential interference contrast (DIC) microscopy image of a representative CAKI-2 cell is shown in Fig. 4(a), with a fluorescence image of the same cell shown in Fig. 4(b). In the DIC image, an arrowhead denotes one of three nucleoli found in the interior of the cell nucleus, and an arrow denotes what is most likely a combination of the Golgi apparatus and endoplasmic reticulum in the perinuclear region [21]. The fluorescence image shows the sequestering of dye within the cell. With the exception of the dark rings at the periphery of the nucleoli, the fluorescence intensity is relatively uniform. Unlike some other common nucleic acid stains (e.g. Propidium Iodide), DRAQ5 exhibits highly specific binding to DNA, and does not require the use of RNase to limit fluorescence from nucleic acid in the cell cytoplasm [19]. Furthermore, DRAQ5 is known to have a low quantum yield [22], making it an ideal contrast agent for PAM. An overlay of the DIC and fluorescence images shown in Fig. 4(c). The overlaid image provides context for the fluorescence; demonstrating the localization of the stain within the nucleus.

 

Fig. 4 a) Differential Interference Contrast image of a CAKI-2 renal carcinoma cell. The arrowhead indicates one of three nucleoli in the nucleus, and the arrow indicates the Golgi apparatus and endoplasmic reticulum. b) Fluorescence image of the DRAQ5 stained nucleus of the cell shown in a). c) An overlay of the images in a) and b) shows localization of the fluorescence signal in the cell. d) PAR image and e) PA image of the cell in a). Both images were generated from data acquired with a single scan. Arrowheads in e) show fluctuations in the PA signal not visible in the fluorescence image. f) The overlay of the PAR and PA images agrees well with the optical microscopy overlay in c). The scale bar is 20 μm and can be applied to all images.

Download Full Size | PDF

PAR and PA images simultaneously acquired from the same cell are shown in Figs. 4(d) and 4(e), respectively. The SNR for the RF-line corresponding to the scan location with the highest PA signal amplitude was 43 dB for PAR, and 27 dB for PA. For the RF-line with the lowest PA signal amplitude, the SNR for the PAR and PA signals were 44 dB and 14 dB, respectively. The PAR and DIC images show similar features, with both the boundary of the cell cytoplasm and the internal structures denoted in Fig. 4(a) readily identified. As cells have negligible absorption in the visible spectrum, structure in the unstained portion of the cell is likely due to a combination of changes in the cell topography and differences in organelle refractive index, e.g. between the nucleus and cytoplasm [23,24]. The PA image is in good agreement with the fluorescence image, with the dark rings at the periphery of the nucleoli readily discernable. A possible reason for these rings is an abundance of rRNA, which is not targeted by DRAQ5 [19], and thus would lead to a decrease in both fluorescent and PA signal. Furthermore, the PA image exhibits nuclear features not present in the standard fluorescence technique (denoted with arrowheads), perhaps indicative of variation in the nuclear chromatin distribution [25]. The composite PAR/PA image is shown in Fig. 4(f), and bears striking similarity to the DIC/fluorescence composite image acquired with the 40X objective. It is prudent to note that the contrast in DIC is primarily due to gradients in the phase difference of light transmitted through the sample [26], and thus, while there appears to be an inversion of contrast in Fig. 4(f) compared to Fig. 4(c), both images provide unique and complimentary information. While the direction of the shadows can be arbitrarily adjusted by adjusting the bias in the DIC image [26], the features of the PAR image are determined by the attenuation properties of the sample, and in the current implementation can only be modified by changing the refractive index of the medium surrounding the cell.

The results of this section demonstrate the feasibility of applying the simultaneous PAR/PA imaging technique to a sample of stained biological cells. Overlaying the PA images on the PAR images provides context for the features observed in the PA image, and does not require co-registration of images acquired using different imaging systems. This is especially advantageous when working with samples such as biological cells, as it can be challenging to locate the same cell when the sample is moved to a secondary system for additional imaging. Furthermore, the composite PA/PAR images provide excellent agreement with well established techniques like DIC/fluorescence, and can be accomplished without the use of CCDs and additional optical components (e.g. condensers, prisms). While the components necessary for traditional light microscopy image formation could potentially be incorporated into existing PAM setups, it would increase both the cost and the complexity of the resultant system. By instead using the PAR technique in PAM systems which lack native optical imaging capabilities, images equivalent to those obtained with traditional optical microscopy can be generated. In the future, we plan to generate label free PAR/PA images of biological cells by leveraging the endogenous absorption of erythrocytes and melanoma cells in the visible wavelengths, and of proteins and nucleic acids in the UV regime [27].

3.4 Triplex imaging of silicon wafer

An optical micrograph of the top surface of the die is shown in Fig. 5(a). The images resulting from the simultaneous PAR/PA/US scan are shown in Figs. 5(b)-5(d). Each image displays unique features dictated by the modality used for image generation. Comparing the simultaneously acquired images, regions which appeared bright in the PAR image (Fig. 5(b)) appeared dark in the PA image (Fig. 5(c)), and vice versa. Due to the low optical absorption of silicon in the near infrared [20], we hypothesize that this is due to the presence of strongly absorbing metallic components, such as aluminum, copper, or tungsten, which are deposited on the wafer surface during fabrication [28]. These metallized layers significantly decrease the detected PAR signal, but are known to yield high PA signals in PAM [29]. However, in addition to being dependent upon the material’s optical absorption properties and thickness, the amplitude of the PAR signal is also dependent upon the laser energy used. Generation of both a PAR and PA signal in these regions may be possible when using higher laser energies. Both PA and US are 3D imaging modalities, and can be used to generate depth-resolved B-Mode images depicting absorber/scatter distribution within the transducer DOF, respectively. In contrast, since the PAR signal is generated by direct optical excitation of the transducer active element, it is not possible to create depth-resolved images with the technique. However, as demonstrated in Fig. 3, the PAR signal amplitude is directly affected by features at any point along the beam path which modulate the laser fluence at the active element. In this way, the 2D PAR images depict features which are well out of range of the transducer’s acoustic DOF, and would otherwise be undetected. For example, diagonal tracks inscribed on the bottom of the wafer during the wafer backgrinding process [30] are readily apparent in Fig. 5(b) (denoted by an arrow), but are not visible in either the PA or US images.

 

Fig. 5 a) Optical image of the top of an integrated circuit die on a silicon wafer substrate. PAR, PA, and US images of the die simultaneously acquired with the triplex technique are shown in b) – d), respectively. Unique features specific to each image are denoted with arrows and arrowheads. b) Arrows in the PAR image show the presence of features inscribed on the bottom surface of the wafer. c) The PA image shows unique features in the metalized regions of the die surface that are not visible in the other modalities. d) The US image shows high contrast at the edges of the different materials (arrowhead), and debris on the surface of the wafer (arrows). The scale bar is 100 μm and can be applied to all images.

Download Full Size | PDF

Intricate surface patterns (denoted with arrowheads) are apparent in the metalized regions of the PA image. While these patterns are also visible in the US image, the contrast is greater with PA, potentially due to the presence of several different metallic components. The aforementioned surface features were not visible in the PAR image (Fig. 5(b)); this is due to the low signal averaging used in this scan. However, as shown in the non-contact PAR scan in Fig. 6, when signal averaging was increased to 150x, the SNR of the PAR signal increased by approximately 23 dB, and similar patterns could be seen in the PAR image of a different die on the same wafer. Future improvements to the system will include the integration of a low-jitter US pulser, allowing for higher signal averaging to achieve high SNR for acquired signals in the triplex scan. Additional features which are unique to the PA image, such as those denoted by the arrows, are likely due to deviations in the thickness or composition of the metal surfaces, or to the quality of the surface bonding [29].

 

Fig. 6 Non-contact PAR scan of a different section of the wafer shown in Fig. 5. The higher signal averaging enhances SNR and makes additional features on the surface of the dye visible. The scan region is 1 mm x 1 mm.

Download Full Size | PDF

High-frequency US images can be used to identify the quality of adhesion between the layers of integrated circuits, as well as identify changes to both surface topography and detect subsurface cracks and voids [31]. Although no such subsurface features were detected in the present triplex US image, high contrast edges were observed at the boundary of the metallized layers on the die surface (arrowhead). The US image is the only image which showed indication of two imperfections on the top surface of the die (denoted by the two arrows in Fig. 5(d), which are also seen in the optical image. These could potentially be debris which have strong optical absorption in the visible regime, but negligible absorption in the near infrared, and are thus not detectable with the present IR setup. Due to the relatively larger size of the acoustic focal spot (approximately 8 μm) compared to the optical modalities, the US image has the poorest lateral resolution of the three images. Nevertheless, fine detail such as the 3 by 3 grids of 10 μm square contacts on the top of the die are still easily resolved.

Acoustic microscopy allows for the detection of subsurface cracks, layer delaminations, and surface topography in a non-destructive manner [7,31]. It can also be used to quantify sample mechanical properties, such as the density, and the speed of sound of longitudinal, shear, and surface acoustic waves [7,31]. In the future, we plan to apply these techniques to the US data acquired with triplex scanning for the purposes of material characterization. Further, while the 3D imaging capabilities of the triplex system were not discussed in great detail in the present work, conventional US B-Mode images can also be generated from the scan data. This makes it possible to gain an appreciation for the depth detail within the sample, as well as the sample surface topography [32]. As the silicon substrate is non-absorbing, such visualizations would not be possible with conventional PAM techniques alone.

4. Conclusion

Our PAR imaging technique derives contrast from spatial variation in a sample’s optical absorption and scattering properties. We demonstrate that in our transmission mode UHF-PAM setup, PAR can be performed simultaneously alongside high-resolution PA and US imaging, and allows for the detection of features well outside the transducer’s DOF. Additionally, we show that the technique can be used to resolve micron-sized objects which exhibit scant optical absorption in both biological and inert samples, provided that a refractive index mismatch between the target and its surroundings exists. Such non-absorbing samples cannot be imaged using traditional PA techniques alone. In cases where only a portion of the sample is capable of generating a PA signal, PAR offers a complementary contrast mechanism that enables visualization of optically scattering structures within the object (e.g. the endoplasmic reticulum in Fig. 4(d)). Thus, PAR provides a means for increasing the amount of information that can be acquired from a sample, and furthermore, broadens the range of samples which can be imaged using transmission mode PAM systems.

However, there are some limitations to the triplex technique. As with other projection-based imaging techniques which have found wide use (e.g. brightfield microscopy, X-ray imaging), PAR is not capable of ascertaining the depth along the optical path at which an inclusion is located. Nevertheless, we believe that PAR can function as an effective tool for the analysis of thin biological samples, such as mouse ears and zebrafish larvae, where PAM and other forms of microscopy can be employed [33]. Currently, the lateral and axial resolutions of our triplex technique are limited by the low NA of the 10X optical objective, and the bandwidth of the acoustic transducer, respectively. Going forward, we plan to use a 0.45 NA objective with a 1 GHz US transducer to push both lateral and axial resolutions to the 1 μm range [11]. At these frequencies, the triplex imaging technique could enable an in-depth, label-free analysis of the optical and material properties of biological cells. Acoustic microscopy could be used to determine the thickness, speed of sound, impedance, and attenuation of biological cells [34,35], while label-free PA imaging would depict the spatial distribution of nucleic acids using UV wavelengths [27] or cytochromes at 422 nm [36]. In such studies, concomitant PAR would reveal the presence of non-absorbing cellular features (e.g. the nucleoli) that would otherwise be undetected. To our knowledge, no other PAM system is capable of acquiring such a diverse amount of information in a single scan. In the future, we plan to use this high-resolution triplex imaging technique to image biological samples, such as zebrafish larvae, and compile co-registered data sets of their endogenous absorption, attenuation, and mechanical properties.