Nano-FTIR vs. PiF-IR: Comparing Nano-IR Techniques

Background

Ever since the invention of the atomic force microscope (AFM), researchers have sought to invent technologies that would bring conventional chemical analysis techniques like infrared spectroscopy to a much smaller spatial volume. Currently, there are a few competing techniques which claim to offer these abilities.

Given the popularity and utility of Fourier transform infrared (FTIR) spectroscopy, one natural option is to extend this technique to the nanoscale via nano-FTIR. However, while FTIR is a robust and user-friendly technique at larger scales, the nanoscale variation has some key limitations that other techniques like photo-induced force infrared (PiF-IR) spectroscopy have alleviated.

Spectroscopy

A chemical spectrum is simply a plot of intensity over wavelength of light. It may sound trivial to describe something so simple; however, it is important to understand that there are two fundamentally different ways to measure a spectrum. One can either use a tunable source, or a tunable detector.

When using a tunable source, the material is illuminated with a very narrow band of wavelength(s). The sample response is measured and then the source is tuned to another frequency. With many data points, a complete spectrum can be created.

When using a tunable detector, the material is illuminated with a broadband source that contains every frequency of light to be measured (this is sometimes called a white-light source). Then, the detector is tuned to only measure light of a certain frequency.

This difference between tunable sources and tunable detectors is one of the fundamental differentiators between PiF-IR and nano-FTIR. This has many implications on how these two nano-IR techniques operate, most of them stemming from tunable detectors having problems from a combination of using an uncontrollable light source and needing very sensitive detection methods.

FTIR

Conventional Fourier transform infrared (FTIR) spectroscopy is a well-established analytical technique that acquires the infrared (IR) spectrum of absorption (or transmission) of a solid, liquid or gas sample. It utilizes a broadband light source so the detection must be tunable. This is accomplished using interferometry, which is not always intuitive to understand.

In an FTIR machine, a sample and a reference mirror form two arms of a Michelson interferometer. As the reference mirror is moved, the two beams from the sample and the reference arm interfere at a photodetector, forming a detector signal versus mirror position graph called an interferogram. A complex Fourier transform is performed on the interferogram to acquire the real (reflection) and imaginary (absorption) IR spectra.

The reason this works to record a spectrum in terms of wavelength is because of the moving reference mirror. As the mirror moves, only a few select wavelengths will coherently interfere with the light from the stationary mirror; all the other wavelengths will destructively interfere producing negligible signal for each mirror position. The response of the sample as a function of mirror position can be used to produce a spectrum by back calculating via a Fourier transform. This will produce a spectrum that shows the sample’s response at each wavenumber. Therefore, while the system is complicated, it is fundamentally a “tunable detector.”

Given the diffraction limit of conventional optics, FTIR is limited to a spatial resolution of ~5 µm. ATR FTIR can do slightly better with a resolution of 3 µm horizontally and a depth of 1.6 µm, but that is still too imprecise when the goal is to look at nanoscale features.

Principle of nano-FTIR

One method to overcome the diffraction limit and achieve higher spatial resolution is to combine FTIR with tapping mode (TM) atomic force microscopy (AFM) to realize nano-FTIR.

Based on an apertureless near-field optical microscope design (also known as scattering scanning near-field optical microscopy, or s-SNOM), nano-FTIR utilizes a modern broadband (white-light) laser source instead of a fixed-wavelength laser as is normally used for s-SNOM imaging. The sample arm of the Michelson interferometer is replaced by the light scattering from the tip-sample interface of the TM-AFM.

In nano-FTIR, the tip is typically metal coated, and the excitation light polarized along the tip direction to exploit the high intensity of tip-enhanced near-field illumination. The near-field signal is measured optically by collecting the light scattered off the tip and using lock-in amplifiers to suppress the far-field signals.

Principle of PiF-IR

Another method to measure nanoscale chemical signatures is via photo-induced force infrared (PiF-IR) spectroscopy. In PiF-IR, a widely tunable narrow band laser is used to excite the sample under an AFM tip. Therefore, PiF-IR is a tunable source technique.

Unlike s-SNOM or nano-FTIR techniques the signal is collected using mechanical force detection rather than the optical detection of scattered light. This means that when comparing nano-FTIR to PiF-IR, there are a few inherent advantages in a PiF-IR system. Specific comparisons are discussed below.

Comparison of nano-FTIR and PiF-IR

Far-field background suppression

Nano-FTIR collects scattered light from the tip to detect the near-field response of the sample due to the excitation light. While effective, this detection scheme comes with a few inherent problems that negatively impact signal strength. Nano-FTIR relies on the fact that lock-in detection of the interference signal at the tapping frequency will mostly reject the far-field background signal. However, the far-field light still contains some components modulated with the tapping frequency. For example, light scattering from the shank of the tip, or light affected by the moving shadow of any part of the cantilever or tip. Therefore, a higher harmonic of the tapping frequency is usually utilized to maximize suppression of the unwanted signal. The cost of using a higher harmonic, however, is reduced signal as each higher harmonic is about a three- to fivefold loss in signal level [1].

As explained in scientific principles, Pi F-IR completely and fundamentally rejects the far-field background signal by its implementation of force measurement, achieving superior SNR.

Spatial resolution

The spatial resolution of nano-FTIR is reported to be approximately equal to the tip radius, which typically is around a few tens of nanometers for metal-coated tips. This contrasts with PiF-IR where an even more confined near-field interaction provides a spatial resolution of ~5 nm for a similarly shaped tip. The relatively poor resolution quoted for nano-FTIR may not be a physical limitation of the technique but may be due to poorer SNR because of the lower power of the broadband laser source (compared to the sharply tuned laser source used for PiF‑IR) and less efficient near-field detection methodology. Nano-FTIR suffers from the same low efficiency of light collection inherent to s-SNOM as described above.

Light sources

Nano-FTIR suffers some inherent power control and signal strength disadvantages due to the broadband white-light source used. As an example, a commercial nano-FTIR utilizes a state-of-the-art broadband mid-IR source which spans 670 to ~2000 cm⁻¹ with an emission bandwidth of about 400 cm⁻¹. This laser source generates an average power level of 1 mW integrated over the bandwidth [2]. If an IR spectrum with 10 cm⁻¹ spectral resolution is desired, roughly 25 µW (or less due to loss from optics) of power is available at each point of the resolved spectrum. This pales when compared to as much as 5 mW of laser power that is available for each ~1 cm⁻¹ bandwidth with the quantum cascade laser (QCL) utilized for PiF‑IR [3]. Per wavenumber, this means a QCL can generate three orders of magnitude higher power than the broadband laser used with nano-FTIR. Even at the typical power of ~ 500 µW utilized for PiFM, the abundance of power allows high SNR spectra with high resolution to be taken far more quickly with PiF-IR (in seconds) than with nano-FTIR (in tens of minutes).

Selective power control

Many nanoscale samples, especially organics and biomolecules, can be easily damaged by high-powered IR light. Therefore, careful power management of the excitation laser is crucial to any nanoscale analytical technique.

With PiF-IR, on samples that can be damaged by excessive intensity and heating, an attenuator is used to reduce the optical power to as little as 0.5% to 10% of the available QCL power. This can be done selectively in real time for any range of wavenumbers which is especially important for wavelength bands where the sample is highly absorptive. This power notching technology avoids sample damage while maintaining high SNR and short acquisition time. This technology can provide massive improvements to the dynamic range of the system­.

Unfortunately, because nano-FTIR uses a broadband light source, this type of power control is impossible. Therefore, usable power levels are constrained by the peak absorption of the sample to avoid damage. In many cases, this means that to get sufficient SNR, multiple spectra must be taken and averaged, further increasing the time required to get meaningful data.

Fixed-wavenumber imaging

Fixed-wavenumber imaging is incredibly useful for mapping chemical variations on the surface of a nano-scale sample. Such images are often especially useful for understanding complex heterogeneous samples, where multiple fixed-wavenumber images taken at different frequencies highlight material components separately in a visually intuitive display. With photo-induced force microscopy (PiFM), the QCL can be tuned to a wavenumber of interest (usually corresponding to a known molecular transition), and then a full image can be made in a matter of minutes.

Since nano-FTIR uses a broadband light source, it cannot acquire fixed wavenumber images directly. Instead, such images must be approximated by taking a hyperspectral dataset and then extracting the intensities from a narrow band of wavenumbers. Unfortunately, this approach is extremely time consuming, generating a full spectrum at each image pixel when only single wavenumber information is needed. While taking a full spectrum at each pixel can be very useful on complex heterogeneous samples with many unknown chemical species, the ideal case is for the user to be able to select between single wavenumber imaging and full spectrum imaging as appropriate. PiF-IR provides both options—single wavenumber imaging, and hyPIR spectra, which are hyperspectral images that provide a full spectrum at each image pixel.

Power normalization and calibration

All good analytic techniques need to have some reference to make sure spectra accurately depict the sample response. Metals such as gold have a flat-IR response, so they can be used as a reference material for calibration. With the broadband laser used for s-SNOM-based nano-FITR, the power profile for a given center wavenumber fluctuates slightly. This necessitates that an interferogram on gold be acquired periodically to normalize the response from the sample [4]. For the most accurate and reliable normalization, the reference interferogram needs to be acquired in identical experimental conditions (tip, harmonic detection, substrate morphology, and other factors) [5].

With PiF-IR, although the tunable laser may also have a nonuniform power output as a function of wavenumber, a reference spectrum is rarely needed. Instead, the power profile across the full spectral range can be made to be constant by an active attenuator. This means that there are fewer opportunities for errors arising from improper power normalization, and without the need for frequent reference spectra, data can be acquired much faster (in as little as 15 seconds for a fully normalized constant-power spectrum or 100ms for a digitally normalized spectrum).

Note: for some very thin samples (less than ~15 nm) differential measurements may be needed to remove substrate contributions. However, this is universally true for all nanoscale molecular characterization.

Thermal stability

Thermal drift can be a significant problem in s-SNOM-based nano-FTIR with the potential of introducing false results or other normalization issues. With a Michelson interferometer, differing thermal expansion in the reference and sample arms causes unwanted phase drift. For a representative commercial s-SNOM system, it is reported that a drift as small as 100 nm of path length difference (between sample and reference arms) shifts the nano-FTIR spectrum by about 6 degrees, which is in the same order of magnitude as the phase shift produced by absorption in many samples [5]. From literature, one can read that enough phase drift takes place in about 120 seconds so that new interferograms of a reference area must be acquired within that time [5]. Other compensation methods can be applied, but phase drift remains a major nuisance for nano-FTIR.

On the other hand, because PiF-IR uses mechanical force detection, drift is hardly an issue at all. For PiF-IR spectra, there is no thermal drift. When taking a PiFM image thermal drift can slowly shift the imaging region as it does in any AFM – an effect which is minimized by good instrument design. As long as precautions are taken to be aware of any residual drift, and the correct location is being imaged, thermal drift has no impact on PiFM or PiF-IR.

Spectral resolution, acquisition time, and SNR

For bulk analysis techniques like FTIR that take spectra from much larger volumes, spectral resolution beyond a certain point doesn’t matter very much. This is because in a larger volume of material there may be molecules in a wide variety of energy states and spatial orientations, resulting in very wide spectral peaks. However, as the volume of probed molecules becomes smaller, there will be fewer molecular states represented. This results in sharper peak shapes and is one of the reasons why spectral resolution matters for nano-scale IR analysis.

PiF-IR uses a QCL with an incredibly narrow spectral resolution of approximately one wavenumber. This means that PiF-IR spectra are extremely detailed to the point where more information can be collected than is possible with conventional FTIR. This opens many possibilities for groundbreaking research. For example, changes in the secondary structures of some proteins can shift a molecular transition by approximately ten wavenumbers, which would be easily detectable in PiF-IR spectra.

While nano-FTIR often has better spectral resolution than many bulk FTIR machines, it can’t match the resolution offered by PiF-IR, a factor which limits its capability in some application spaces. Due to the interferometer design, higher spectral resolution requires more time for spectral acquisition. For nano-FTIR, the length that the reference mirror travels to generate the interferogram determines the spectral resolution. The longer the travel, the higher the spectral resolution. The speed of the travel governs the SNR. Examples from published literature suggest a single pass of the mirror with a travel range sufficient for 8.3 cm⁻¹ spectral resolution takes around 40 seconds to cover a bandwidth 200 to 400 cm⁻¹ wide. Unfortunately, due to poor signal strength, multiple passes may be needed for sufficient SNR. Furthermore, at this resolution, it is likely necessary to interrupt the measurement multiple times for reference spectra because of thermal drift in the interferometer. This means the whole process for a high-resolution spectrum can take as much as an hour or more for good SNR if the wavelength range is large. Therefore, the types of high-resolution spectra that can be recorded in seconds using PiF-IR become impractical to replicate using nano-FTIR. In general, calculations suggest that PiF-IR measurements are between 100 to 1000 times faster than comparable nano-FTIR measurements.

Summary

	nano-FTIR	PiF-IR
Laser source	Broadband (~350 cm−1)	Narrowband (~1 cm−1)
Near-field signal	Scattered photons mixed with far-field background photons	Photo-induced force only under the tip apex
Background suppression	Lock-in detection of higher harmonics	No background signal
Spectrum technique	Fourier transform of interferograms	Wavenumber sweep
Spectral resolution	6.4 cm−1 (optional 3cm−1)	~1 cm−1
Signal to noise	Poor	Excellent
Speed	Slow, especially for high-resolution	Very fast
Need for reference	Yes	No
Selective power control (notching)	No	Yes
Spatial resolution	~20 nm	less than 10 nm
Agreement with FTIR	Good	Excellent
Fixed-wavenumber imaging	No, only via hyperspectral imaging	Yes

PiF-IR and nano-FTIR offer two different methods of probing surface chemistry at the nanoscale. While each claims to offer similar features, the detection schemes and light sources make a huge difference in how effective each technology actually is. PiF-IR relies on a mechanical detection scheme that fundamentally eliminates background signals from contaminating the near-field response while nano-FTIR collects scattered light and uses lock-in detection to filter competing background signals at the cost of signal strength. Additionally, PiF-IR uses a high-powered narrow-band tunable laser which offers much greater control and signal strength than the slower broadband white-light sources used in nano-FTIR experiments. These key differences mean that PiF-IR based measurements have tremendous advantages in terms of better signal strength, faster data acquisition, higher thermal stability, and precise laser power management. Photo-induced force microscopy and spectroscopy techniques are therefore the most valuable nanoscale characterization techniques for all researchers and technicians.

References

1. S. Amarie and F. Keilmann, “Broadband-infrared assessment of phonon resonance in scattering-type near-field microscopy,” Phys. Rev. B 83, 045404 (2011)
2. Toptica FemtoFiber dichro MidIR Brochure
3. QCL Brochure
4. M. Autore, L. Mester, M. Goikoetxea, and R. Hillenbrand, “Substrate Matters: Surface-Polariton Enhanced Infrared Nanospectroscopy of Molecular Vibrations,” Nano Lett. 19, 8066 (2019)
5. I. Amenabar, S. Poly, M. Goikoetxea, W. Nuansing, P. Lasch & R. Hillenbrand, “Hyperspectral infrared nanoimaging of organic samples based on Fourier transform infrared nanospectroscopy,” Nature Communications 8, Article #: 14402 (2017)