Overcoming Artifacts: Polarized Raman of Individual Electrospun Fibers Under Immersion

Introduction

In recent decades, materials scientists have reported striking enhancements in the mechanical,^1–3 thermal,^4,5 electrical,^6,7 and optical ⁸ properties of electrospun fibers as their diameter decreases, with the most remarkable properties typically observed in fibers below 500 nm. As a result, electrospun nanofibers have become prime candidates for a range of applications, including filtration,^9,10 coatings,^11,12 flexible electronics, ¹³ sensing, ¹⁴ and tissue engineering.^15,16

Understanding the structural basis of this trend is key to developing rational fabrication strategies and optimizing properties for scale-up. Polarized confocal Raman microscopy has emerged as a powerful technique for probing the structure of individual fibers of various compositions, owing to its high signal-to-noise ratio, sensitivity to both amorphous and crystalline phases, reasonably rapid analysis time, and non-invasive nature. ¹⁷ Notably, robust Raman methodologies have been developed to quantify molecular orientation (i.e., chain alignment) in single fibers using sets of four spectra collected under different light polarization conditions (hereafter referred to as “standard methodologies”).^18,19 Raman has helped reveal a strong correlation between property enhancement and molecular orientation, which exhibit an analogous increase as fiber diameter decreases.^20,21 This orientation improvement aligns with electrospinning dynamics, where thinner fibers experience stronger elongational forces and faster solvent evaporation—conditions that favor greater and more sustained chain alignment.^22–24 More recently, we have applied polarized Raman microscopy to elucidate the process-structure relationships in individual fibers. Our results show that parameters such as polymer crystallinity ²⁵ and glass transition temperature (T_g), ²⁶ solvent volatility ²⁶ and affinity with the polymer, ²⁷ and type of fiber collector ²⁸ influence the degree of molecular orientation across a wide range of fiber diameters, ultimately enabling the production of fibers with enhanced and more uniform performance.

Yet, the characterization of ultrathin fibers remains analytically challenging. While investigating molecular orientation in individual polyacrylonitrile (PAN) fibers using polarized Raman microscopy, Papkov et al. observed an unexpected inversion of the polarization contrast for fibers below 500 nm in diameter, suggesting a preferential chain orientation perpendicular to the fiber. ²¹ This contradicted results for larger PAN textile fibers and was inconsistent with the general understanding of the electrospinning process.^20–24 The inversion was therefore attributed to artifacts stemming from both instrumental limitations and fiber geometry, which can skew molecular orientation analysis by polarized Raman microscopy^21,29,30 and other techniques, such as infrared (IR) spectroscopy. ³¹ As detailed in a subsequent perspective, ¹⁷ instrumental issues can arise from the polarization-dependent response of the system and from motion of the fiber during the measurement or the switching of polarization. These problems are exacerbated for fibers with a diameter smaller than the diffraction limit of the laser, where slight shifts can displace the focus completely off the fiber surface. These instrumental effects can generally be mitigated with optimized optics and spectrum validation protocols to detect abnormal signal loss. ¹⁷ Spectral distortions intrinsic to the fiber geometry relate to differences in reflectivity and transmittivity when the fiber is probed under different polarizations. These distortions are more pronounced for thinner fiber due to the combined effect of enhanced birefringence (caused by higher molecular orientation) and greater fiber curvature. The latter causes most incident light to strike the fiber at oblique angles, where reflectivity varies for light polarized parallel and perpendicular to the fiber axis, the most extreme case being at the Brewster angle. Collectively, these effects cause one polarization intensity to appear artificially stronger than the other, easily overshadowing spectral variations from molecular orientation and thus compromising accurate quantification in ultrathin fibers using standard methodologies. ¹⁷

To circumvent these artifacts, Papkov et al. applied a semi-quantitative method that requires a single polarized spectrum by calculating the intensity ratio of two bands with distinct polarization directions. ²¹ The band ratio does not quantify the order parameter directly, but it is proportional to molecular orientation. While effective for PAN, this approach is not directly applicable to polymers whose vibrational modes are less well characterized ²⁶ or for which the intense modes are all aligned in the same direction. ²⁵ A more quantitative strategy is to build a calibration curve that relates a band intensity ratio (measured under a single polarization) to the quantitative order parameters determined on oriented macroscopic samples using standard methodologies.^20,28,30 While effective in some cases, challenges include the capability to prepare bulk calibration standards covering a wide range of order parameters, ²⁵ the presence of bands whose intensity ratio correlates well with orientation, ²⁶ and absence of confounding spectral overlap in multicomponent systems. ²⁷ This underscores the need of experimental strategies for the reliable investigation of molecular orientation in fibers with diameters close to or below 500 nm.

Since the artifacts appear to be due to the refractive index (n) mismatch at the interface between the cylindrical sample (n often ∼1.5 for organic materials) and the surrounding medium (typically air with n = 1), reducing this index mismatch should attenuate reflection at the fiber interface and enable signal collection that better reflect molecular orientation. In fact, performing measurements in an index-matching immersion medium is a common practice in optical microscopy and its use in Raman microscopy has been strongly recommended by Everall,^32,33 and Everall et al., ³⁴ although often overlooked, to minimize aberrations and to enable accurate depth profiling and subsurface imaging. To our knowledge, such an index-matching strategy has not yet been explored to mitigate polarization-related artifacts in molecular orientation measurements.

This work aims to expand the application range of standard Raman orientation quantification methodologies to small nanofibers by eliminating geometry-induced artifacts. More specifically, we investigate the molecular orientation in PAN electrospun fibers of a wide range of diameters in three “immersion” media, i.e., air, deionized water, and silicone oil. While standard methodologies in air replicate the large distortions previously reported, immersion in deionized water attenuates these artifacts, and oil “immersion” provides order parameters that closely match those obtained from a single-polarization calibration method. Thus, oil immersion allows the reliable application of standard methodologies for investigating challenging electrospun systems composed of specialty polymers or blends.

Experimental

Results and Discussion

Polyacrylonitrile (PAN) is a semicrystalline polymer of considerable practical importance, valued for its high mechanical strength, chemical resistance, and thermal stability and its role as a precursor to carbon fibers.^38,39 Electrospun PAN fibers provide an excellent model system for evaluating the effectiveness of higher-refractive-index immersion media in mitigating artifacts related to fiber geometry in polarized confocal Raman microscopy. PAN fibers are typically electrospun from dimethylformamide (DMF) or dimethyl sulfoxide (DMSO), ³⁹ low-volatility, highly polar solvents that promote the formation of ultrafine fibers,^22,40,41 associated with exacerbated geometric artifacts. Both their Raman spectra and the associated artifacts are well documented.^17,35,42

Figure 1 presents the Raman spectra polarized parallel (ZZ) and perpendicular (XX) to the sample main axis for three PAN fibers of comparable diameters (∼800 nm) interrogated in air, deionized water, and silicone oil, as well as for a stretched PAN film in air. The 2244 cm^–1 band, assigned to the nitrile stretching mode, ³⁵ was used as the molecular orientation marker because it provides a high signal-to-noise ratio under all conditions and is well isolated from other PAN and immersion media bands (see Figure S1 in Supplemental Material for a comparison of the Raman spectra of PAN, deionized water, and silicone oil). The main component of its Raman tensor is estimated to form a tilt angle of ∼70° with respect to the chain axis.^35,36 Its intensity should therefore be higher under the XX polarization and lower under the ZZ polarization when PAN chain orientation increases. This expected behavior is observed for the representative film of Figure 1, associated with an order parameter ⟨P₂⟩ of 0.25, which does not show geometric artifacts typical of fibers thanks to its flat geometry. In sharp contrast, the fiber measured in air (n = 1.00) displays a marked inversion of the nitrile band intensities, with a stronger intensity in the ZZ spectrum than in the XX spectrum. This inversion is consistent with that identified in previous polarized Raman studies on single PAN fibers and incorrectly suggests a strong polymer chain alignment perpendicular to the fiber axis.^17,21 Further evidence of artifacts is provided by the stronger XX intensity of the 1458 cm^–1 band, associated with –CH₂– bending, ⁴³ which is largely insensitive to molecular orientation, as indicated by its equivalent intensities under XX and ZZ polarizations in the oriented film spectra. Interestingly, a similar loss of intensity equivalence for an orientation-insensitive band was also observed in our previous study on poly(ethylene terephthalate) (PET) fibers, when their diameter was reduced below 500 nm. ³⁰

OPEN IN VIEWER

Figure 1.

Polarized Raman bands of PAN used for orientation quantification, measured on three representative individual fibers in air (“Air”), deionized water (“Water”), silicone oil (“Oil”), and on an oriented film in air. All samples exhibit a 〈P₂〉 of ∼0.25, determined using either the standard methodology (film) or a calibration curve (fibers; vide infra). For easier comparison, the spectral intensities are normalized to the 1458 cm^–1 band under ZZ polarization.

Under deionized water (n = 1.33) immersion, the 2244 cm^–1 band shows comparable XX and ZZ intensities. While this suggests a rather isotropic chain distribution, which is more plausible than the result in air, it remains very unlikely as PAN fibers electrospun under similar conditions typically display molecular anisotropy and enhanced mechanical properties along the fiber axis, even at ∼800 nm diameter.^1,21 This spectral behavior is instead attributed to a partial mitigation of the geometry-induced artifacts by reducing the refractive index mismatch at the fiber/medium interface. A possible confounding factor is the change in numerical aperture from 0.8 to 1.0 when switching from the 100× to the 60× objective. A few control experiments were therefore conducted with the 60× objective under dry conditions, which led to the return of the incorrect relative intensities between the ZZ and XX spectra. The quality of focus was also affected, and the spectral intensity was reduced to only ∼15% of that obtained with the 100× objective. We attribute these problems to severe spherical aberrations due to the incorrect use (in air) of a water immersion objective with numerical aperture (NA) = 1.0. Finally, Figure 1 shows that when the 100× objective is used to measure the spectra of fibers covered by a layer of silicone oil (n = 1.403), the polarized intensities of the nitrile band recover a ratio similar to that of the oriented film, confirming the preferential molecular orientation along the fiber axis and, importantly, a strong suppression of geometric artifacts.

To establish the quantitative effectiveness of high refractive index media to mitigate the artifacts during the polarized interrogation of fibers, we must compare the apparent ⟨P₂⟩ values and the true ⟨P₂⟩ values obtained using an artifact-free method. We previously demonstrated that a calibration curve relating ⟨P₂⟩ values to a band intensity ratio under a single polarization provides a reliable means of determining the orientation for fibers with sub-micrometer diameters measured in air.^20,28,30

Figure 2 presents the calibration curve developed for this study relating ⟨P₂⟩ values of stretched PAN films, calculated from sets of four polarized spectra using the standard methodologies, to the 2244 cm^-1/1458 cm^–1 band intensity ratio under ZZ polarization, where the 1458 cm^–1 band serves as an internal reference. The linear regression of the calibration data demonstrates a strong correlation (R² of 0.97) between the intensity ratio and molecular orientation. This curve was used to determine the artifact-free ⟨P₂⟩ values of all PAN fibers investigated, solely from their ZZ spectrum acquired during an initial air measurement prior to immersion. Notably, it confirmed that the ⟨P₂⟩ values of the three fibers whose polarized spectra are shown in Figure 1 are approximately 0.25, comparable to the film value, indicating that the variations in their polarized spectra arise from geometry-induced artifacts. We emphasize that, as not all polymers feature an appropriate orientation-insensitive band, the development of experimental strategies that enable the reliable quantification of molecular orientation in ultrafine fibers using standard methodologies is necessary.

Figure 3 compares ⟨P₂⟩ values obtained using the standard methodologies (⟨P₂⟩_SM) with ⟨P₂⟩ values derived from the calibration curve (⟨P₂⟩_calib) for each fiber investigated in air, deionized water, or silicone oil. The black slope represents the “ideal” scenario where ⟨P₂⟩_SM consistently matches ⟨P₂⟩_calib. Fidelity scores, indicated in the figure, assess how closely each dataset meets this scenario (details on the calculation of the score are available in the Supplemental Material). Fibers measured in air consistently display negative ⟨P₂⟩_SM values that diverge significantly from their ⟨P₂⟩_calib values, which range from 0.01 to 0.48. Even worse, about 50% of ⟨P₂⟩_SM values, in the shaded gray region, fall below the lowest physically meaningful ⟨P₂⟩ value of –0.5. This dataset yielded a fidelity score of 0%, indicating no overlap between ⟨P₂⟩_SM and ⟨P₂⟩_calib. These results are consistent with the dominance of geometry-related artifacts in the sets of polarized spectra used to quantify ⟨P₂⟩_SM for fibers measured in air. Under water immersion, fibers still show discrepancies between their ⟨P₂⟩_SM, which range from –0.25 to 0.10, and their ⟨P₂⟩_calib, which range from 0.01 to 0.40. Nonetheless, the average difference between the two values is much smaller than for fibers measured in air, and the fidelity score increases to 68%, reflecting a substantial reduction of the artifacts. However, fewer fibers were successfully analyzed in water than in air due to sporadic movement during acquisition, leading to loss of focus and inconsistent intensities. This results in a selection bias in favor of larger fibers of lower orientation, leading to an overestimated fidelity score. Finally, fibers interrogated in silicone oil show a close agreement between their ⟨P₂⟩_SM and ⟨P₂⟩_calib, with overlapping ranges from –0.04 to 0.70 for ⟨P₂⟩_SM and 0.09 to 0.57 for ⟨P₂⟩_calib, and a commendable fidelity score of 85%. Interestingly, the high viscosity of the silicone oil minimized fiber movement compared to the water immersion measurements, such that the selection bias for larger fibers does not play a large role.

OPEN IN VIEWER

Figure 2.

Calibration curve for evaluating the artifact-free 〈P₂〉 of PAN fibers by correlating the 2244 cm^-1/1458 cm^-1 intensity ratio in ZZ spectra with the 〈P₂〉 parameter. The curve was established with 31 PAN films stretched to various draw ratios and analyzed using the standard MPD methodology.

OPEN IN VIEWER

Figure 3.

Comparison between 〈P₂〉 values obtained using the standard methodologies (⟨P₂⟩_SM) and ⟨P₂⟩ values derived from the calibration curve (⟨P₂⟩_calib) for fibers interrogated in air, deionized water, and silicone oil. For each fiber, ⟨P₂⟩_calib was determined from the ZZ spectra measured in air. The black slope represents the expected trend for perfect agreement between ⟨P₂⟩_SM and ⟨P₂⟩_calib. The percentages indicated on the figure reflect the fidelity of the datasets (color-matched) to this trend. The shaded grey region encloses data points associated with ⟨P₂⟩_SM values below –0.5, which are physically impossible.

Figure 4 presents the dependence of 〈P₂〉_SM values of fibers measured in oil and 〈P₂〉_calib values on fiber diameter, a key trend for the optimization of electrospun materials.^20,21,25 To highlight the main conclusions, data are represented as median values, with marks indicating orientation dispersion from the 25th to 75th percentiles. Both datasets exhibit a typical increase in molecular orientation with decreasing diameter, with median 〈P₂〉_SM and 〈P₂〉_calib values reaching 0.37 and 0.40, respectively, for diameters below 600 nm. An onset diameter of orientation increase appears in both cases around 800 nm, consistent with previous studies on PAN fibers. ²¹ Interestingly, a substantial molecular orientation is retained in fibers with a diameter above this onset, with median 〈P₂〉_SM and 〈P₂〉_calib values of 0.21 and 0.19, respectively. This behavior can be attributed to the crystalline phase of PAN whose orientation relaxation is restricted, as previously observed for fibers of highly crystalline polymers. ²⁵ We note that some fibers in the window below 600 nm may have diameters under 400 nm. Indeed, scanning electron microscopy (Figure S2 of the Supplemental Material) shows that electrospinning at the lowest concentration of 10% m/V mostly leads to fibers with diameters between 400 and 600 nm, but also to some outliers of lower or higher diameter. Our spectral method for diameter determination (see the Experimental section) is limited by the laser diffraction limit and would therefore overestimate the diameter of these ultrafine fibers.

OPEN IN VIEWER

Figure 4.

Diameter dependence of molecular orientation derived from 〈P₂〉_SM of fibers measured in oil “Oil (SM)” and 〈P₂〉_calib “Calib”. The symbols represent the median of data points within 200 nm diameter windows, and the vertical marks delimit the orientation dispersion from the 25th to 75th percentiles for each window. The first data point includes fibers thinner than the laser spot for which the reported diameter may be overestimated.

The modest differences in median 〈P₂〉 observed in Figure 4, along with the 85% fidelity score for the 〈P₂〉_SM of fibers measured in oil (Figure 3), indicate that immersion in oil avoids most geometry-induced artifacts. Indeed, the inherent errors in Raman measurements due to drifting, slight loss of focus, and uncertainties related to spectral baselines and band intensities, typically introduce an inaccuracy of ± 0.05 in the 〈P₂〉 values for both the standard methodologies and calibration curve (error bars are omitted in Figure 4 for clarity). This uncertainty is contained by optimizing the measurement time to strike a balance between increasing the signal-to-noise ratio and lowering the risk of signal loss. Further improvement can be achieved on the instrumental side by using a high-performance anti-vibration table, an automated polarization-switching device, and minimizing air flow near the fibers. When appropriate measures are taken, immersion in an index-matching medium stands out as an effective strategy to mitigate or eliminate the polarization-dependent reflection artifacts due to the size and curvature of fibers with diameters close to the laser diffraction limit.

Conclusion

This work introduces a practical approach for minimizing spectral artifacts caused by the cylindrical geometry of electrospun micro- and nanofibers, which can impede the accurate determination of molecular orientation using conventional polarized Raman spectroscopy methodologies. By immersing electrospun fibers of polyacrylonitrile (PAN) in deionized water (n = 1.33) or silicone oil (n = 1.403) to reduce the refractive index mismatch with the surrounding medium (typically air, n = 1), it is possible to reduce or almost eliminate the geometry-induced artifacts that affect the intensity of polarized Raman spectra. The accuracy of the quantitative orientation data was significantly improved, with 〈P₂〉 values obtained under water and oil immersion displaying fidelities of 68% and 85%, respectively, relative to calibrated reference data, compared to a fidelity of 0% for fibers interrogated in air. The accuracy achieved under oil immersion further enabled us to establish a robust 〈P₂〉-diameter curve, a useful approach to optimize the structure-properties relationships in electrospun materials. This refractive index-matching strategy is an effective and relatively easy to implement solution to determine accurate Raman orientation values for challenging samples such as thin electrospun nanofibers.

Supplemental Material