Julian
Hniopek
abc,
Josefine
Meurer
de,
Stefan
Zechel
de,
Michael
Schmitt
bc,
Martin D.
Hager
*de and
Jürgen
Popp
*abce
aDepartment Spectroscopy & Imaging, Leibniz Institute of Photonic Technology, Albert-Einstein-Str. 9, 0775 Jena, Germany. E-mail: juergen.popp@leibniz-ipht.de
bInstitute of Physical Chemistry (IPC), Friedrich Schiller University Jena, Helmholtzweg 4, 07743 Jena, Germany
cAbbe Center of Photonics, Friedrich Schiller University Jena, Albert-Einstein-Str. 6, 07745 Jena, Germany
dLaboratory of Organic and Macromolecular Chemistry (IOMC), Friedrich Schiller University Jena, Humboldtstr. 10, 07743 Jena, Germany. E-mail: martin.hager@uni-jena.de
eJena Center of Soft Matter (JCSM), Friedrich Schiller University Jena, Philosophenweg 7, 07743 Jena, Germany
First published on 6th June 2023
Stimuli-responsive polymers can switch specific physical properties in response to a change of the environmental conditions. This behavior offers unique advantages in applications where adaptive materials are needed. To tune the properties of stimuli-responsive polymers, a detailed understanding of the relationship between the applied stimulus and changes in molecular structure as well as the relationship between the latter and macroscopic properties is required, which until now has required laborious methods. Here, we present a straightforward way to investigate the progressing trigger, the change of the chemical composition of the polymer and the macroscopic properties simultaneously. Thereby, the response behavior of the reversible polymer is studied in situ with molecular sensitivity and spatial as well as temporal resolution utilizing Raman micro-spectroscopy. Combined with two-dimensional correlation analysis (2DCOS), this method reveals the stimuli-response on a molecular level and determines the sequence of changes and the diffusion rate inside the polymer. Due to the label-free and non-invasive approach, it is furthermore possible to combine this method with the investigation of macroscopic properties revealing the response of the polymer to the external stimulus on both the molecular and the macroscopic level.
Ideally, these substrates can be designed by polymeric materials featuring a reversible structural unit. The most commonly applied reversible groups are azobenzenes and spiropyrans,10,13,14 which can simply be switched by light irradiation. Furthermore, ferrocene or disulfide containing polymers can be applied for switching the wettability by reversible redox reaction.15,16
To understand and optimize the molecular properties required for the desired switching behavior, it would be desirable to directly monitor the molecular structure of the responsive materials during the application of the external stimulus, which leads to a change of the properties of the material.
However, currently in most cases, the change of molecular structure can only be correlated indirectly via studying the change of certain properties, e.g., wettability determined by contact angle measurements.17 This technique allows the analysis of the interaction of a substrate with water (or oil) over time and during stimulus application. However, a correlation of the molecular change and the interaction of the substrate with the water (oil) droplet is only possible with laborious methods, such as introducing fluorescent particles into the structure.18
A less complex method to monitor the structural changes alongside the macroscopic changes in surface properties would therefore be beneficial in order to optimize such intelligent substrates further and to understand structure–property relationships in more detail. We propose Raman micro-spectroscopy as a powerful tool to monitor the chemical structure of responsive polymers in situ during stimulus application. It is important to note that this technique enables us to measure simultaneously the stimulus progression within the material and the response of the chemical structure of the polymer, and to correlate this with the changed properties (cf.Fig. 1b). Raman spectroscopy is an established tool for the analysis of smart polymers and has been applied, for example, to monitor the structural changes in electro-responsive,19 shape-memory,20 and self-healing polymers,21,22 as well as to monitor chemical processes like diffusion23 or dynamic bond opening in polymers in general.24 Here, the opportunity to combine Raman microspectroscopy with two-dimensional correlation analysis (2DCOS), introduced by Isao Noda,25 makes it possible to mathematically retrieve relationships between stimulus progression and molecular changes inside the polymer. This enables the derivation of structure–property relationships in a straightforward and unbiased manner. Furthermore, Raman spectroscopy offers the distinct advantage that it directly probes the molecular structure of the sample via its characteristic molecular vibrations and is thus label-free.26 As a consequence, there is no need for expensive and laborious modifications of the polymers to enable the observation of structural changes, which moreover might influence the properties of the polymer. Furthermore, water does not show strong Raman signals, due to its highly polar nature and resulting comparatively low scattering cross-section.27,28 Therefore, Raman spectroscopy can be combined with methods like contact angle measurements or water-based triggers (e.g., changes of the pH-value) in a straightforward manner without significant contribution of the water to the spectrum.29,30
Wavenumber position/cm−1 | Assignment |
---|---|
1600 |
ν(C![]() |
1656 |
ν(C![]() |
1702 |
ν(C![]() |
2110, 2180, 2250, 2280 | ν(C–D), d3-AcOD |
Due to the FT based detection approach and the use of an excitation wavelength relatively far in the near infrared region, FT-Raman is ideally suited to acquire reference spectra free of fluorescence with very high spectral precision. It can therefore serve as a gold standard tool to identify artifacts introduced by fluorescence or other optical effects in the following confocal Raman microscopy measurements. To acquire the spectra, powdered polymer P3 was placed in an aluminium pot and a drop of heavy water/deuterated acetic acid (D2O/d3-AcOD, 9:
1 v/v) was added. Due to the higher atomic mass of deuterium compared to protium, deuterated species lead to a characteristic shift of the molecular vibrations involving hydrogen atoms (C–D/O–D) to lower wavenumbers. This shift is especially pronounced for the respective stretching vibrations, which are shifted from over 3000 cm−1 to the “silent” region of the spectrum (2200–2500 cm−1). In this spectral region, only very few moieties show vibrational bands, which enables straightforward separation of the signals arising from the activating mixture (D2O/d3-AcOD) and the signals of the polymer.
Besides the obvious appearance of the d3-AcOD related signals in this silent region, adding the activating mixture also changes the spectral region under 1750 cm−1 (decreasing the intensity at 1656 cm−1 and increasing the intensity at 1702 cm−1). These bands do not correspond with bands of the d3-AcOD and must, therefore, stem from changes of the polymer upon addition of the stimulus, i.e., the acid. Indeed, the positions of these bands correspond well to the CN stretching vibration of an imine (1656 cm−1), and the C
O stretching vibration of an aldehyde (1702 cm−1),27 indicating the stimulus-response of the imine bond by hydrolysing the imine to an aldehyde. Together with the well separated signals of the d3-AcOD, this enables orthogonal investigation of the presence of acetic acid (the stimulus) and the onset of molecular changes within the polymer (the response).
Afterwards, a drop of a 9:
1 (v/v) mixture of D2O/d3-AcOD was placed directly on top of the focused points and the scan was started to afford a spatially and temporally resolved data set. The right panel of Fig. 1c exemplarily shows a snippet of the time series recorded 12.5 μm below the surface. As determined via the reference measurements (cf.Fig. 1c, left), hydrolyzation of the imine (decreasing the intensity at 1650 cm−1 and increasing the intensity at 1700 cm−1), as well as the influx of the deuterated acetic acid (appearance of bands between 2100 and 2300 cm−1) can be seen. Naturally, as the reaction progresses, both the degree of hydrolyzation (response) and the amount of deuterated species (progressing stimulus) increase with time. Looking more closely at the traces, the influx of solvent and the hydrolyzation seem to happen on slightly different timescales, as hydrolysed polymer bands can be seen already before acetic acid bands can be detected (first signals at 60 s). This indicates that the reaction mainly happens on the interface between the activating mixture and the polymer. Consequently, before switching the polymer, only a minute amount of the activation mixture (and therefore d3-AcOD) is present inside the focal volume of the confocal Raman microscope. While this minute amount is large enough to cause hydrolyzation of the polymer, its concentration stays below the detection limit of the measurement setup. After the hydrolyzation, the polymer changes from being more hydrophobic (caused by the butyl groups of the imine) to being hydrophilic (free aldehyde), which allows the bulk of the (water-based) activation mixture to penetrate deeper into the polymer.
As a method with molecular specificity, Raman spectroscopy enables the amount of hydrolyzed imines to be monitored over the activation process.
For this, the band area of the imine was integrated and compared with the band area of a C–C backbone vibration (1220 cm−1), revealing a constant level during the FT-Raman reference measurements. By monitoring this ratio with respect to the ratio between the two bands in a pristine sample (AImine/AC–C = 1.63), the hydrolyzation ratio can be calculated. Here, for all depths a hydrolyzation ratio of 65–81% could be found before the aforementioned effect of polymer swelling set in (for calculation details see the ESI†). To further investigate the changes in the data, we employed two-dimensional correlation analysis (2DCOS), which has been described in detail elsewhere.25,37,38 In short, 2DCOS recovers the cross-correlation function between spectral variables (bands) under the external perturbation, thereby revealing information about their relationship (direction and especially order of changes). Since 2DCOS uses a well-defined mathematical approach to get this information it is less susceptible to subjective interpretation as, e.g., plotting kinetic traces or difference spectra and can also deal with more complex response (e.g., not strictly decreasing) functions that are not easily analyzed visually. 2DCOS is a well-established method to analyze sets of spectra recorded under external perturbation (in this case time after addition of the activation mixture) in general,39–42 and in the study of smart polymers in particular.21,22 2DCOS results are usually displayed in the form of contour plots, that depict the real (synchronous, Φ) and imaginary (asynchronous, Ψ) parts of the complex correlation function. These maps visualize if spectral changes within a data set are correlated (Φ) and if a phase-shift is present between the changes (Ψ), respectively. Rules for interpreting these maps can be derived from the underlying maths, the so-called Noda Rules.38
(1) A positive correlation signal in Φ shows that changes at the respective x and y positions are correlated and they change in the same direction.
(2) A negative correlation signal in Φ shows that changes at the respective x and y positions are correlated and they change in different directions.
(3) If the signs of the signals in Φ and Ψ are the same for a combination of spectral positions x and y, the change at x happens before the change at y.
(4) If the signs of the signals in Φ and Ψ are different for a combination of spectral positions x and y, the change at y happens before the change at x.
A plot with the synchronous (Φ) and asynchronous (Ψ) parts of the 2DCOS at 25 μm depth can be seen in Fig. 3. It can be derived that the bands of acetic acid (2100–2200 cm−1) are positively correlated (positive signal in Φ → 1st Noda Rule, change in the same direction) with the signal for the hydrolyzed polymer at 1702 cm−1 and negatively correlated (negative signal in Φ → change in different directions, 2nd Noda Rule) with the CN signal of the imine signal in the unhydrolyzed polymer. This once again confirms the observation that addition of the activation mixture causes the polymer to hydrolyze. Furthermore, by combining the signals in Φ and Ψ the temporal relationship of the changes to the polymer structure and the influx of acetic acid can be elucidated.
To interpret the signals in Ψ the 3rd and 4th Noda Rules must be applied, which state that the change at the respective wavenumber position plotted on the x-axis happens before the change at the y-position if the signs of a signal are the same in Φ and Ψ and vice versa. Since the signs of the signal at (1702, 2110), (1656, 2110) and (1656, 2180) cm−1 are all equal in Φ and Ψ, it is confirmed that the diffusion of acetic acid always happens after the polymer has been hydrolyzed. The same plots, focused on this area of interest for all depths can be found in the ESI (Fig. S7–S11†) showing the same behavior in all depths. Additionally, in Fig. 3, correlation signals with the aromatic vibration at 1600 cm−1 can be seen, which stand in for generally observable correlation signals with all polymer bands outside of the shown wavenumber region. These correlation signals are caused by the aforementioned swelling of the polymer and subsequent loss of Raman intensity after the liquid has diffused into the polymer. As can be seen from the differing signs of all correlation signals along the (1600, y) cm−1 line in Φ and Ψ these changes are strictly happening later than the hydrolyzation as well as the diffusion of liquid.
The power of 2DCOS to extract small, systematic changes from a series of spectra could be utilized to directly visualize the correlation of the aldehyde/imine bonds with the influx of D2O as well. For this purpose, the O–D stretching vibration located between 2500 and 2600 cm−1 could be used. While this band shows very low Raman intensity, which makes integrating the band impossible, using 2DCOS it was still possible to find correlations with the CO/N stretching bands (cf. Fig. S12–S16 in the ESI†). These maps show similar results as shown in Fig. 3 and S7–S11.† While the correlation intensities are lower and the asynchronous correlation is therefore only clearly visible for the aldehyde C
O stretching vibration at 1702 cm−1, they also indicate bulk influx of the D2O only after hydrolyzation of the imine. Together, these 2DCOS observations further support the mechanism proposed in the previous section: the hydrolyzation reaction is mediated by minute amounts of the activation mixture without significant penetration of the imine-containing polymer. Only after the opening of the imine bond and a resulting change in hydrophilicity can the majority of the mixture penetrate the polymer, leading to the observed lags in their Raman signals.
Furthermore, recovering the correlation function between two signals allows for a direct way to measure the delay between the two processes. The average phase angle between the signals in the Fourier domain is encoded in the 2DCOS, which directly relates to the delay between the two signals in the time domain.43 Using the parameters of the Fourier Transform, it is possible to transform those phase angles back to a value in the time domain, which reveals delay times between hydrolyzation and influx of acetic acid ranging from 30 to 55 s (cf.Table 2, for details on the calculation see the ESI†). There is no increase or decrease of the delay time with increasing depth, indicating statistical fluctuations around a constant delay time between hydrolyzation and acetic acid influx. This delay time is governed by the diffusion rate of the activation mixture in the switched polymer, which can be calculated as 0.27–0.3 μm s−1 from the delay times and spacing of the measurement points. Consequently, the diffusion coefficients are in the range of (16.9–18.9) × 10−9 cm2 s−1.
Depth | 0 μm | 12.5 μm | 25 μm | 37.5 μm | 50 μm |
---|---|---|---|---|---|
a Correlating molecular behavior to macroscopic properties. | |||||
Delay time/s | 43 ± 4Exp | 46 ± 4Exp | 42 ± 4Exp | 41 ± 4Exp | 44 ± 5 |
Diffusion rate/nm s−1 | 293 ± 3Exp | 270 ± 3Exp | 301 ± 3Exp | 303 ± 3Exp | 282 ± 3 |
Diffusion coefficient/10−9 cm2 s−1 | 18.3 ± 1.4Exp | 16.9 ± 1.4Exp | 18.8 ± 1.4Exp | 18.9 ± 1.4Exp | 17.6 ± 1.6 |
Comparing these values to the literature values obtained for the diffusion of acetic acid in an epoxy resin (rate: 37 nm s−1)44 and pure PMMA (coefficient: 1.2–3.2 × 10−9 cm2 s−1 depending on the diffusion model),45 these values are roughly an order of magnitude higher. This fits the expectation, as the diffusion of a very polar electrolyte is measured and free aldehydes are more polar compared to the polyether structures present in epoxy resins as well as the ester moieties in PMMA, thereby underpinning the validity of the measurement.
As a vibrational spectroscopic method, Raman micro-spectroscopy is broadly applicable to many functional groups and thus presents a technique transferable to different stimuli-responsive systems. The technique is in principle only limited by the acquisition speed and the separability of signals of the activating species and polymer. The technique applied here, isotope labelling, is also broadly applicable and in the case of deuterium labelling usually comparatively easy and cheap to carry out. Using non-linear Raman techniques or resonance enhancement, it is furthermore possible to increase the acquisition speeds to the order of milliseconds, thus enabling the use of the same basic technique also for faster processes.
All chemicals were used as received from TCI (Eschborn, Germany), Sigma Aldrich (Darmstadt, Germany), Alfa Aesar (Kandel, Germany), Thermo Fisher Scientific (Geel, Belgium) and Acros Organics (Geel, Belgium) if not otherwise stated. All solvents were dried over a molecular sieve under a nitrogen atmosphere. The stabilizer in the used liquid monomer methyl methacrylate (MMA) was removed over a short aluminium oxide (AlOx) column (neutral AlOx, obtained from Molecula, Darlington, UK).
Nuclear magnetic resonance (NMR) spectra were measured using a Bruker AC 300 (300 MHz) spectrometer (Billerica, MA, USA) at 298 K if not stated differently. The chemical shift is given in parts per million (ppm on δ scale) related to deuterated solvent.
Size exclusion chromatography (SEC) measurements were performed using the following setup: a Shimadzu instrument with a CBM-20A (system controller), DGU-14A (degasser), LC-20AD (pump), SIL-20AHT (auto sampler), CTO-10AC vp (oven), SPD-20A (UV detector), RID-10A (RI detector), PSS SDV guard/1000 Å/1000000 Å (5 μm particle size) chloroform/isopropanol/triethyl-amine [94/2/4] at 1 mL min−1 at 40 °C, and poly(methyl methacrylate) (standard).
Raman-spectroscopic measurements were performed on a MultiRAM near-infrared Fourier-transform Raman spectrometer (Bruker Corporation, Billerica, Massachusetts, United States of America) in the range between 100 and 4000 cm−1 with a spectral resolution of 4 cm−1. The Raman excitation light at 1064 nm was provided by a Nd:YAG laser (Klastech DeniCAFC-LC-3/40, Dortmund, Germany). The laser power was set to 1000 mW at the sample plane and 128 scans were accumulated for each sample.
Raman microscopy was performed on an alpha300 Raman microscope (WiTec GmbH, Ulm, Germany) equipped with a 600 line per mm grating. The excitation source was a continuous-wave diode-pumped solid-state laser (CoboltTM, Solna, Sweden) at 514 nm and 30 mW power in the sample plane. Light was focused onto the sample using a Leica Fluotar long-working distance objective (100×/0.6 NA). Spectra were recorded in the wavenumber region between 0 and 4000 cm−1. A depth scan with 2 × 5 points spread 10 μm laterally and 50 μm axially was performed using an integration time of 2 s per point (for an illustration of the measurement scheme, see Fig. S1 in the ESI†).
All pre-processing of the Raman spectroscopic data and subsequent analysis, including 2D correlation analysis, were performed using GNU R, version 4.2.2 with in-house developed scripts.46 A detailed description of the analysis steps can be found in the ESI.†
Contact angle measurements were performed on an OCA 30 (DataPhysics, Filderstadt, Germany) using the SCA 20 software in dynamic tracking mode.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3sc01455j |
This journal is © The Royal Society of Chemistry 2023 |