Polyvinyl acetate-based polymer host for optical and far-infrared spectroscopy of individualized nanoparticles

A wide variety of nanoparticle (NP) systems, ranging from colloidal nanocrystals to single-wall carbon nanotubes (SWCNTs) to nanoplatelets, are synthesized using solvent-based techniques. Although solutions (aqueous, inorganic, or organic) are usually the preferred environments for these nanomaterials, researchers often require their NPs to be in a solid form for additional measurements, integration into a device or apparatus, and/or technological implementation. When NPs are intentionally transitioned from the solution state to a solidified form, say, for example, for spectroscopic measurements, certain key considerations are required. For optical spectroscopy, in particular, the NP host or the substrate must show some degree of transparency, stability over a reasonable time duration, and resistance to damage from experimental conditions (light, temperature, magnetic field, etc.). These conditions can often be fairly easily met with the appropriate forethought and the material selection process. However, additional requirements are often needed beyond these bare-minimum constraints. When the NP host goes from a solution environment to a gel or solid, the properties of the NPs are often altered due to factors such as NP aggregation,¹ host-created strain,² unwanted environmental coupling, and increased optical scattering.³ These unwanted behaviors add another, more stringent, level of host-material constraints, especially since the desired NP properties are often masked by these effects.

Given the premium that researchers and technical facilitators have placed on using NPs in a solid-state form, numerous techniques to incorporate solution-based NPs into films, solids, and/or solid-state inclusions have proliferated. In addition to simple drop casting and ink printing,^4,5 researchers use transmembrane vacuum filtration,^6–11 inclusion into hydrogels,¹² polymer-exchange processes,¹³ dry pressing,¹⁴ spin-coating solutions across substrates,^15,16 shearing NPs between substrates,¹⁷ dip¹⁸ and spray coating,^19–22 and deposition or synthesis in a polymer matrix.^23–35 Many of these techniques require multiple processing steps, considerable time, specialized equipment, and an assortment of chemicals, which makes them very challenging to implement and use by non-specialists. Additionally, the use of dry pressing, spin/dip/spray coating, and inclusion into different hydrogel or polymer matrix hosts often adds a significant amount of NP aggregation and/or strain.^1,23,36 Furthermore, it is often the case that the best non-NP host/substrate medium is not ideal for particular optical regimes, such as the low-energy (long-wavelength) spectrum, where transparencies in the mid-infrared (⁠$≈5$ $μ$m to 20 $μ$m) and/or far-infrared/terahertz (THz) (30–3000 $μ$m or 10–0.1 THz) regions are required.

In this work, we present a rapid, facile, low-cost method for producing free-standing, flexible, and robust polyvinyl acetate (PVAc-)-based films that have low absorption in both optical and THz regimes. Using a combination of empirical and machine-learning techniques, we find the optimal drying conditions to reliably produce optical quality samples. Additionally, we show that one- and zero-dimensional aqueous-based NPs can be included into these polymer matrices without significant NP aggregation or changes in their optical response. Specifically, we demonstrate that minimal strain is created in SWCNTs and Au NPs when immersed in these PVAc-based films. Taken together, we reveal a straightforward method for optical spectroscopists to create high-quality, low scatter films for measurements that cover a broad electromagnetic spectrum.

The base polymer matrix for these films is a commercially available PVAc-based polymer solution (Clear Elmer’s Glue$TM$⁠).³⁷ Aliquots of this solution were either drop cast onto glass coverslips or spin coat (Laurell WS-400B-GNPP/LITE) at 100 rpm (10.5 rad s⁻¹) for 40 minutes under controlled air flow and temperature, $T$⁠. As discussed later, we used a variety of air flow rates, temperatures, NP mass loading, and drying times to produce optical quality polymer films. Typically, the highest film uniformity was achieved using a spincoater but drop casting also created excellent films. Once dried, polymer samples were peeled off of the coverslip with a razorblade and used as free-standing structures, a process illustrated in Fig. 1(a).

To create NP-immersed PVAc films, we vortexed mixtures of 60% NP solution and 40% PVAc solution (by mass fraction) together in 1.5 ml centrifuge tubes. In addition to this optimized 60:40 NP-to-PVAc solution ratio, we also found other ratios near this split could also produce high-quality films.

Unsorted SWCNTs produced using the CoMoCat process were purchased as raw powder from Sigma Aldrich [Carbon Nanotube, single-walled, (6,5) chirality, batch No. MKBS9733V]; this soot is commonly described as SG65i material. Sorted SWCNTs were enriched from a different CoMoCat process source soot obtained directly from Southwest Nanotechnologies (batch No. SG76-L24). SWCNT powders were tip sonicated (Branson Sonifier 250) using a $14inch$ (6.35 mm) microtip at 10 W  ml⁻¹ nominal input power for 4.5 h (0.145 W ml⁻¹–0.19 W  ml⁻¹ transferred) for 10 ml–15 ml of 1 mg ml⁻¹ SWCNTs in 2% DOC.³⁸ From these solutions, enriched SWCNTs were made using aqueous two-phase extraction (ATPE).^39–42 Along with the preparation of individualized SWCNT solutions for our immersed NPs, we also used citrate-capped 40 nm diameter Au NPs in an aqueous solution, which were purchased from Nanocomposix.

Several spectroscopic techniques were used in this work, including optical absorption (Varian Cary 500 Scan spectrophotometer), Fourier-transform infrared spectroscopy (FTIR) (Bruker Tensor 27), THz time-domain spectroscopy (TTDS), Raman spectroscopy, and photoluminescence excitation spectroscopy (PLE), which were all performed on polymer films with and without NPs. To ascertain the pulsed laser-induced damage threshold (LIDT) of the PVAc-based polymer film, we focused an amplified, 4 mJ/pulse, $≈200$ fs, 800 nm Ti:S laser (Libra from Coherent) down to a measured beam waist of 260 $μ$m. For the continuous wave (CW) LIDT measurements, a 3.4 W, 4 MHz linewidth, tunable-wavelength Ti:S laser system (Matisse from Spectra-Physics) set at 765 nm was focused to a measured beam waist of 264 $μ$m.

In addition to experimental observations, we used numerical modeling to understand the role of index of refraction on our Au NP spectra. Specifically, the Au NP plasmonic absorption peaks were compared to a Mie scattering model using the PyMieScatt Python library. For this calculation, we assumed that all of the gold spheres were the same size. Mie scattering efficiencies were calculated from 450 nm to 659 nm using a wavelength-dependent complex index of refraction for Au.⁴³ 

We note at the outset that PVAc is a polymer that has been previously used in thin-film spectroscopy, though its optical properties have not been as comprehensively measured as in this work.^32–34 Many of these PVAc environments have been rendered transparent through the addition of benzene and other processing steps.^32,34 Another polymer material commonly used for optical and THz spectroscopy is polyvinyl alcohol (PVA). Similar to the previously mentioned benzene processing of PVAc, PVA films are frequently made with extensive processing steps that our technique does not require^28–31 making our system much more amenable for quick and effective film preparation. Finally, in addition to this work, other researchers have also used off-the-shelf commercially transparent PVAc-based glues due to their optimized durability, drying properties, and transparencies, as well as their ease of use.³³ 

Given that we are evaluating a polymer matrix for multiple types of optical spectroscopy, we examined several standard methods of probing NPs. A prime example is Raman scattering spectroscopy in which the frequency difference between incident and scattered light is measured to provide insights into phonon energies. In Fig. 1(b), we show the Raman spectrum for the PVAc-only dried polymer film. A distinct feature around 2900 cm⁻¹, which we attribute to C—H bonds, is observed. Importantly, the negligibly small magnitude of the polymer Raman spectrum does not interfere with the much larger NP Raman signatures.

We also investigate the applicability of our PVAc films for transmission-based optical measurements. In Fig. 1(c), we present the optical absorptivity spectrum, $α(λ)$⁠, of the PVAc-only film from 0.2 $μ$m to 5000 $μ$m. Using a combination of optical absorption, FTIR spectroscopy, and TTDS, we can see that two broad regions of high transparency are clearly discernible. The first region, from $≈300$ nm to 2800 nm, shows a low $α$ below 5 mm⁻¹. This key regime extends from the UV through the near-infrared, which makes these polymer matrices especially applicable for biological and medical samples, as well as condensed matter NPs that have important properties in the visible or near-infrared regimes. After that, we observe strong absorption peaks between 2.5 $μ$m and 27 $μ$m, which are consistent with bond-stretching peaks determined in PVAc by Ahmed.⁴⁴ 

The second region of transparency is one that is very often ignored, especially for aqueous-based materials. From $≈124$ $μ$m to 5000 $μ$m, we find that the polymer absorptivity is only $≈5$ mm⁻¹. While this absorptivity magnitude is not as low as polymers like polytetrafluoroethylene (PTFE; Teflon), high-density polyethylene (HDPE), and polypropylene, which are roughly an order of magnitude lower, it is comparable to polyacrylonitrile-butadiens-styrene (ABS) and polyethylene terephthalate (PET).⁴⁵ With the increasing use of TTDS for low-energy spectroscopy of condensed matter NPs and bio-medical systems,^46,47 having a matrix that is transparent in this region is critical. Moreover, having a NP host system that is optically transparent in both the optical and THz regimes allows for the study of optically created effects that can be probed as a function of time (and frequency) using TTDS as the system returns to equilibrium.^48–51 Though techniques needing both optical and far-infrared wavelengths are not in abundant use, several advanced methods do require such material constraints. Select crystals, such as z-cut quartz, have dual optical and far-infrared transparency windows, but are not amenable to hosting NPs, except when deposited on the surface. Beyond that, very few other materials that can host NPs satisfy this criteria. As such, having this aqueous polymer matrix represents an important advance for hosting NPs relevant to visible, near-infrared, and far-infrared spectroscopists.

As already discussed, the low and flat absorptive region that begins around 125 $μ$m and continues to 5000 $μ$m (the extent of our TTDS bandwidth) represents a significant area of THz transparency, which is surprising in a polymer due to molecular rotational and vibrational modes. Figure 2(a) shows the TTDS time traces for the air reference and PVAc polymer film. Since we are measuring the actual THz electric field with this technique, a transformation into the frequency domain [Fig. 2(b)] allows us to calculate the real and imaginary parts of the electric permittivity, $ϵfilm′$ and $ϵfilm″$⁠, respectively, for the PVAc-based polymer. Figure 2(c) shows $ϵfilm′$ and $ϵfilm″$ for our films from 200 GHz to 2.5 THz. Although $ϵfilm′$ is close in value to other polymers, such as PET, polyaramid, polymethlmethacrylate (PMMA), and HDPE, it varies from 4 to 2 within a short frequency range. As a comparison, across a similar region, PET varies from 3.0 to 2.7.⁴⁵ Similarly, we see a high degree of variability in $ϵfilm″$⁠, although this value is much more in line with other transparent polymers. We suspect that inconsistencies in the thickness across the film (higher at the edges and lower in the middle) may account for some of the variability. Unlike other optical measurements, the beam spot sizes in TTDS vary strongly with frequency due to the order of magnitude span of wavelengths in the far-infrared covered by the pulse. If smaller frequencies probe, on average, thicker regions of the polymer film while larger frequencies measure thinner regions, then this effect could help account for the steep negative slope of both $ϵfilm′$ and $ϵfilm″$⁠. However, that still would not change the overall conclusions shown in the figure but rather would suggest that the true value of $ϵfilm′$ and $ϵfilm″$ for the probed spectral regime lies around 5 meV–6 meV (⁠$≈1.2$ THz to 1.5 THz).

Given these encouraging optical results, we next investigated the environmental conditions necessary to reproducibly make high-quality films. Depending on the drying parameters used, the structure of the dried PVAc-based films varied considerably. To qualitatively categorize our films for use in spectroscopic measurements, we defined a 1 cm diameter region of interest (ROI) at the center of the film. Either SWCNT NPs or green dye was included into the polymer films to clearly indicate regions of inhomogeneity and particulate clumping. Three assessment categories were defined for determining the suitability of a given film for spectroscopic work: “highest optical quality,” “high optical quality,” and “low optical quality.” We defined a “highest optical quality” film as one that has a highly homogenous NP dispersion that is also free of any defects. Here, we define defects as bubbles, differential tension drying rings, thickness variations, and visible sample concentration variations. When a film was considered to be “high optical quality,” the ROI was free of defects but they existed elsewhere in the film. If defects were located in the ROI or if they were spaced such that a 1 cm diameter defect-free region could not be located in the sample, the film was considered “low optical quality;” for this latter result, the films are not suitable for use in optical experiments.

We explored NP loading (fraction of the film composed of NPs), polymer-to-solution ratios, spin-coating speed, spin-coating duration, film diameter, temperature, and applied drying air flow to determine the conditions that produced PVAc-based films devoid of defects. Several of these parameters, such as NP loading and the polymer-to-solution ratios, were quickly established and were consequently held fixed, while others we changed to optimize. Figure 3 shows a plot of qualitative description of films as a function of temperature, air flow, and post-dropcast or -spin-coat drying duration. Films that either had a 25 mm or 50 mm diameter were included in this assessment.

To aid with understanding this drying parameter space, we use a machine-learning algorithm to predict values of temperature, applied air flow, and drying duration that would likely result in, at minimum, “high optical quality” films. Our machine-learning algorithm is based on a k-nearest neighbors (KNN) algorithm was applied to the given data set. As a classification learning algorithm, where predictions are based on the statistical assumption that similar outcomes lie within a given distance (Euclidean) away from each other, KNN lends itself nicely to the given parameter space and classifiers. After a trained model was created based on the given data set, we fed permutations of different parameters into the model and extracted predictions of a “high optical quality” film.

Along with actual data points, Fig. 3 shows (in green) the KKN-predicted parameter space where “high optical quality” films can be produced. These regions, which like the actual data points are projected along 2D planes, help provide strong drying parameter estimates for other researchers using our film-making procedure. Probably the most significant prediction is that somewhat high (25 $°$C to 75 $°$C) temperatures coupled with a strong (⁠$≈25$ ml s⁻¹ to 35 ml s⁻¹) air flow are expected to yield the best drying conditions. A reasonable place to start experiments, however, would be 40 $°$C and $≈25$ ml s⁻¹. Although we have found empirically several high-quality films outside of this sweet spot, as well as several low quality films inside of it, we maintain that by and large, the best set of drying parameters is in this range. We note that defects are often formed when one or more small bubbles make their way into the film as it dries. The qualitative nature of our assessment means that the placement and number of these small bubbles can easily change our film description. Thus, a region of drying parameters that is most likely to produce “high optical quality” films will always produce some “low optical quality” films. This unavoidable spread in bubble placement, however, does not undermine our confidence in claiming that our method is reproducible.

A key condition for using any type of host for pulsed optical spectroscopy is the ability to withstand high intensities (power per unit area) and fluence (energy per unit pulse per unit area). The use of polymers in pulsed laser experiments have traditionally been avoided due to the the low thermal conductivities, heat capacities, and structure damage thresholds of polymers relative to hard condensed matter systems. To assess the suitability of our PVAc-based polymer for intense field work, we performed laser-induced damage threshold (LIDT) measurements with pulsed and CW lasers.

To determine the pulsed LIDT of our polymer films, we used a 200 fs amplified pulsed laser with a central wavelength of 800 nm and a repetition rate of 1 kHz. For both pulsed and CW LIDT measurements, we measured the power of the transmitted light both with and without the polymer film. The pulsed LIDT measurements revealed that at a peak intensity 0.29 TW cm⁻², we began to observe damage to the polymer. In contrast, the CW LIDT experiment showed that damage occurred at an average intensity of 2.2 TW cm⁻². However, we note that this LIDT value is an averaged quantity due to spatial variations created by microscopic inhomogeneities in the film. To put these values for this PVAc-based polymer into context, the CW LIDT for Ge is $≈0.3$ MW cm⁻²,⁵² while modern femtosecond beamline mirrors have a pulsed LIDT of 2.5 TW cm⁻². Our LIDT values are comparable to these systems only because the absorptivity of our films for the UV-visible-IR region is so low: by absorbing very little light, the process of optically induced thermal strain leading to material damage is greatly reduced.

In addition to the LIDT experiments, we also immersed these films into liquid nitrogen, and separately, brought them down to 5 K. In both cases, due to the negligible water remaining in the films after drying, these polymers (both with and without NPs) showed that they could survive cryogenic freeze-thaw cycling without any structural damage. Unlike aqueous-based polymeric hosts and certain crystals, these PVAc-based polymer films can be used for low temperature work.

We created SWCNT-polymer films to determine the ability of this polymer system to host 1D NPs without significant aggregation or induced strain; an example of a SWCNT-polymer film is shown in the inset of Fig. 1(c). To measure the local environment of SWCNTs, we relied on optical measurements of their vibration properties, absorption features, and emission characteristics, which are well known to be excellent determinants of strain, aggregation (bundling), and optical/electronic proximity.

Figure 4(a) shows the Raman spectra from both the SWCNT starting solution and SWCNTs from that solution in a dried PVAc-based polymer film. Importantly, all of the Raman spectroscopic signatures of the SWCNTs, including the radial-breathing mode (RBM), G and D peaks, and 2D peak,^54–56 are clearly maintained when the SWCNT solution is incorporated into the polymer host. To determine if the incorporated SWCNTs are subjected to any polymer-induced strain, we compared the 2D Raman feature from SWCNTs in solution to those in the film. As seen in Fig. 4(b), the shift between these two SWCNT environment conditions is barely visible. From the change in the Raman 2D peak position between film and solution environments of $≈0.4cm−1±0.5cm−1$⁠, we estimate a possible polymer-induced strain on SWCNTs of $≈0.03%±0.03%$ from known relationships between Raman shifts and strain.⁵³ We note for clarity that we are defining error here as the Gaussian width obtained from our spectral fitting. While this strain value is low compared to other strain measurements on SWCNTs, the uncertainty of the small change in the peak position is comparable (if not larger) than the actual change, suggesting that within our ability to measure, the strain-induced peak shift is insignificant.

Absorption peaks in the visible to the near-infrared region, shown in Fig. 5(a), correspond to exciton (optically created, Coulombically bound electron-hole pairs) absorption features in SWCNTs. Given their relationship to both the SWCNT band structure and the nanotube dielectric environment, the position and linewidth of these absorption features are sensitive to the immediate environment of the SWCNTS.^20,57–61 Although all of the absorption features observed in the parent solution are indeed preserved in the film with minimal broadening and minor peak shifts, our best confirmation of SWCNT individualization (i.e., minimized aggregation) is through PLE. For SWCNT chirality identification, researchers use the unique combination of an excited exciton absorptive transition (typically done at optical frequencies) and emissive transition (usually measured in the near-infrared).^54–56 A SWCNT excitation that produces multiple emissions, instead of a single one, corresponds to either a closely associated chirality, a phonon-assisted transition, or an energetic coupling between different chiralities that is indicative of aggregation (bundling).¹³ 

Figure 5(b) compares PLE measurements of sorted semiconducting SWCNTs in the solution and in a PVAc-based film. The transition of the solution-state SWCNTs to a film-based solid shows minimal additional emissions from different coupled SWCNTs. The small degree of increased inter-chirality SWCNT coupling suggests that although nanotube bundling has increased in the PVAc-based film, it is still relatively minimal compared to previous SWCNT film examples.^{6,7,9–11,28,29,62,63}

To extend the applicability of these PVAc-based films, we examined including citrate-capped Au NPs into this polymer host while still preserving their most important optical properties. Based on the specified NP diameter of 40 nm, we expected a plasmonic scattering peak at 524 nm from estimates via Mie scattering calculations of Au NPs surrounded by water (⁠$n=1.33$⁠). This estimate is very close to the measured plasmonic peak in the solution of 522 nm, as seen in Fig. 6. After incorporating these NPs into a PVAc-based film, the NP scattering peak shifted by 18 nm (⁠$≈80$ meV) to 540 nm. In addition to the peak position shift, the Gaussian width was broadened from 21 nm to 29 nm.

As before, the shift in the scattering peak value to a higher wavelength is in excellent agreement with Mie scattering predictions: we used exactly the same NP parameters as for the solution case, except now, we have changed the surrounding index of refraction to 1.484 to account for the polymer host. Plasmonic scattering features from individualized NPs are known to be dependent on the indices of refraction of both the particles themselves, as well as the surrounding medium,⁶⁴ which means that this matching of experimental and computational results through an index of refraction substitution provides three key pieces of information: first, a confirmation that the majority of the observed peak shift is due to NP medium optical effects; second, an implicit confirmation of the continued individualization of the Au NPs even in a polymer host; and third, an estimate of the index of refraction for our PVAc-based polymer in this wavelength regime. The agreement between our experimental observations and model calculations shown in Fig. 6 strongly suggests that the change in the index of refraction of the surrounding medium (water to PVAc) is responsible for the scattering peak shift.

This conclusion does not imply that polymer-created strain is not present, but rather, it is a minimal effect on the optical properties of these NPs, similar to our observations with SWCNTs. We note that more direct methods for measuring and understanding NP agglomeration than we used here, such as small-angle scattering methods (e.g., small-angle neutron scattering, dynamic light scattering, and small-angle x-ray scattering)⁶⁵ and electron microscopy (e.g., multiple thin slices imaged by transmission electron microscopy), are available. However, while these methods measure the inter-nanoparticle spacing and NP aggregate size distribution and average density, they are highly complex and challenging to correctly use, which suggests the need for further work on this PVAc-based polymer.

In addition to water, we also tried to create polymer films using both polar (methanol, ethanol, and acetonitrile) and non-polar solvents (hexanes, toluene, and chloroform). As with the films made with water, each solvent was mixed and vortexed with the PVAc-based polymer as described earlier. At this juncture, separation was observed when hexanes and toluene were used, while chloroform, methanol, and acetonitrile created an opaque, congealed mass. Only films made with ethanol produced transparent, stable films, albeit never comparable in quality to those made with water.

We used a commercial PVAc-based polymer to create thin, transparent films that could host NPs of a different dimensionality, specifically SWCNTs and Au NPs, without significant aggregation. Based off our spectroscopic findings and the fact that we believe that the polymer agglomeration mechanism in this system is not strongly based on the NP electronic properties, it is likely that this polymer matrix can also be used for semiconducting and insulating (e.g., nitride and oxide) NPs. In addition to studying polymer-created aggregation, we also showed that the complex permittivity from 0.2 THz to 2.5 THz is similar to other polymers. Unlike other NP polymer hosts, this PVAc system can be easily cast with and without NPs into large (2.5 cm–5.0 cm) films with high transparency in both the optical and THz regimes. Our work enables other researchers to quickly and efficiently produce high-quality NP-polymer films for optical spectroscopic work in a variety of environments.

H. V. Wladkowski and W. D. Rice thank the University of Wyoming School of Energy Resources for support. This work was authored by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding for J. L. Blackburn provided by the U.S. DOE, Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences, Solar Photochemistry Program. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. J. A. Fagan was supported by National Institute of Standards and Technology internal funds.

The data that support the findings of this study are available from the corresponding author upon reasonable request.