A new 3D microscopy tool in the geologist’s kit

In 1999 Sunney Xie and coworkers at Pacific Northwest National Laboratory published an article containing an image of three cervical cancer cells, in which each cell’s mitochondria appear as bright yellow splotches in a sea of red cytoplasm. ¹ The researchers used no fluorescent labels or stains to generate the color contrast; rather, they used a technique known as coherent anti-Stokes Raman scattering (CARS) microscopy.

The technique was developed in the early 1980s at the US Naval Research Laboratory, but it went virtually unnoticed until Xie and colleagues rediscovered and refined it. In the past decade, CARS has stirred considerable interest among biologists, and now work by Robert Burruss (US Geological Survey), Aaron Slepkov (Trent University, Peterborough, Ontario), Albert Stolow, and Adrian Pegoraro (both at National Research Council Canada) suggests the technique could also prove useful to geoscientists. ² Applied to translucent samples of sedimentary rock, CARS microscopy produced richly informative three-dimensional maps that could shed new light on geochemical and geophysical processes.

Coherent Raman scattering

In CARS microscopy, chemical contrast is generated with a pair of laser pulses—a pump pulse and a Stokes pulse—whose frequencies ω_p and ω_S differ by the frequency ω_v of a molecular vibration of interest. As depicted in figure 1a, a pump photon excites the molecule to a virtual state and a Stokes photon stimulates an emission that leaves the molecule in a vibrationally excited state. When a second pump photon scatters off the vibrating molecule, it can emerge as a so-called anti-Stokes photon, which has blueshifted frequency ω_aS = ω_p + ω_v. By mapping the intensity of the blueshifted signal as a function of location in the sample, one can construct a 3D map of the molecule’s spatial distribution.

Raman scattering is notoriously inefficient; it typically occurs just once in every million or so scattering events. Thus conventional, spontaneous Raman scattering techniques yield a frequency-shifted signal that can be difficult to distinguish from noise. However, if Raman scattering is stimulated with lasers, as it is in CARS, it yields a coherent, and therefore much more intense, signal.

Speedy CARS

To implement CARS using Xie and colleagues’ method, one has to synchronize a pair of picosecond pulsed lasers. Then, to scan a spectrum of vibrational resonances, one must repeatedly retune the lasers’ frequencies. Three years ago, Stolow and coworkers at the National Research Council Canada devised a simpler, quicker strategy that involves just a single, femtosecond pulsed laser. ³

As sketched in figure 1b, they split a laser beam into two arms, one of which is then redshifted—by way of a photonic crystal fiber and bandpass filter—for use as the lower-frequency Stokes beam. As with any laser, however, a decrease in pulse duration comes at the cost of an increase in bandwidth. So the femtosecond pulses should be thought of not as having specific frequencies ω_p and ω_S but as containing broad distributions of frequencies that are centered on those values. In fact, the pulse bandwidths can be tens of times broader than the spectral lines they are meant to detect. That presents a seeming conundrum: How does one tune the difference frequency of two pulses to achieve a spectral resolution exceeding that of the pulses themselves?

The trick was to use what are commonly known as chirped pulses. Using blocks of dispersive glass, each pulse is stretched in a way that temporally sorts its light according to frequency: A miniature observer who watches the chirped pulse go by would see its frequency grow linearly with time. Stolow likens it to a “piano run” in which every note is played in sequence, from low to high. To understand how the chirp facilitates fine-tuning of two pulses’ difference frequency, imagine someone playing two piano runs, one with each hand. As long as each hand progresses at the same pace, it’s possible to maintain a constant spacing of just a few keys between hands.

Analogously, Stolow and company fix the pulses’ chirp rates so that a miniature observer who watches them pass would see a constant, and potentially quite narrow, frequency difference. That difference can then be retuned over a swath of spectrum by adjusting a mirror to alter the time delay between pulses; conveniently, there’s no need to retune the lasers themselves.

Rocks in 3D

Even the most pristine mineral samples are flecked with tiny pockets of trapped liquid or gas known as fluid inclusions. Sometimes just microns in size and containing femtomoles of material, those inclusions help paint a picture of how Earth evolved over geological time scales; each inclusion amounts to a snapshot of the physical and chemical state of a particular part of Earth’s crust at a particular moment in time. Interpreting the snapshot calls for identifying the inclusion’s contents and its formation history; inclusions can be trapped when a mineral precipitates out of the liquid phase, or they can form later on when a mineral fractures and heals.

On learning of Stolow’s CARS microscopy technique, Burruss, a geochemist, thought it might be useful for imaging methane inside fluid inclusions. In theory, methane can be detected with conventional Raman scattering, but when an inclusion also contains long-chain hydrocarbons, the larger molecules’ natural fluorescence tends to obscure methane’s Raman spectral lines. That’s particularly troubling for petroleum scientists, who need to know how methane is—and was—distributed in Earth’s crust in order to unravel the processes responsible for generating fossil fuels.

Applying CARS microscopy to translucent samples of quartz, calcite, and hornblende, Burruss, Stolow, and company found they could easily identify methane and several other chemical constituents of microscopic inclusions, even when long-chain hydrocarbons were present. What’s more, the intense femtosecond pulses also give rise to second harmonic generation (SHG): At interfaces within the material, occasionally two photons of frequency ω_p are converted to one photon of frequency 2ω_p. By mapping spatial variations in that frequency-doubled signal, the team could construct images of a sample’s internal microstructure.

Figure 2 shows four SHG images of a quartz sample superimposed on a conventional microscope image. The SHG images reveal inclusion walls, microfractures, grain boundaries, and other microstructures. Paired with CARS, an SHG image allows one to determine not just the contents of an inclusion, but also whether the inclusion is associated with a particular fracture event. For example, the composite SHG–CARS image in figure 2 shows methane-rich inclusions embedded along a healed microfracture.

Burruss envisions several new applications for CARS microscopy, including visualizing how different minerals and crystalline domains intermingle inside rocks. Broadly speaking, he draws an analogy with atomic force microscopy: “When geoscientists first found out about AFM, they said, ‘Wow! This could be really useful for learning about minerals,’ and now AFM techniques are used extensively in the geosciences. I think there’s a similar opportunity with CARS.”