Confinement of thermoresponsive microgels into fibres via colloidal electrospinning: experimental and statistical analysis

Abstract

The strategy of confining stimuli-responsive microgels in electrospun fibres would allow the fabrication of polymeric networks that combine the microgels swelling ability and properties with the interest features of the electrospun fibres. Colloidal electrospinning is an emerging method in which fibres containing microgels can be produced by a single-nozzle and designed through the solution carrier materials. The incorporation of poly(N-isopropylacrylamide) (PNIPAAM) and PNIPAAM–chitosan (PNIPAAM–CS) in poly(ethyleneoxyde) (PEO) fibres via colloidal electrospinning producing composite fibres was the main purpose of the present work, which was confirmed by means of Scanning Electron Microscopy (SEM). Dynamic light scattering was used to analyse the microgels hydrodynamic diameter ranging up to 900 nm depending on the composition and temperature of the surrounding medium. By performing a statistical analysis the relationship of the processing variables over the fibre size was evaluated following the response surface methodology (RSM). From the set of parameters aimed to minimize the fibre diameter, composite fibres with an average diameter of 63 nm were produced. Only the as-prepared microgels with higher monodispersity provided “bead-on-a-string” morphologies.

---

Introduction

In the field of polymeric materials, microgels – intramolecularly crosslinked polymer particles – are found to be in an outstanding position due to their versatility from both fabrication/synthesis and application perspectives.^1,2 For instance, stimuli responsive or “smart” polymeric microgels can be synthesised by radical polymerization of intrinsic thermoresponsive monomers^3–6 or copolymerized with monomers that could infer additional functionalities.^7–9 They can also be hybridized by encapsulation or in situ growth of metallic/magnetic nanoparticles embedded in the microgel matrix ending up with multi-responsive systems.^10–12 By exploiting such versatility, microgel particles can be fabricated for a wide spectra of applications ranging from industrial (paints, ink printings, cements, oil recovery) to promising biomedical applications² (protein purification,¹³ drug delivery,¹⁴ molecular imprinting,¹⁵ carriers,¹⁶ among others). It is of importance to highlight that microgels have also been thought of as building blocks to infer self-healing properties in layer-by-layer fabricated hydrogels.¹⁷ Authors took advantage of their swelling ability and dynamic in the solvated state to induce a fast and autonomous self-healing ability to the hydrogel.^18,19

Following this thread, in this contribution we pursue the confinement of thermoresponsive microgels in electrospun fibres that could behave as active sites in electrospun non-woven mats, as a first step toward the fabrication of a polymeric network that would have potential biomedical applications and added self-healing abilities. In particular, we propose the colloidal electrospinning method as a tool for both the confinement of thermoresponsive microgels and the development of a desired architecture/morphology. In this line, encapsulating microgels in a fibre matrix is a process mainly dictated by the features and behaviour of the precursor materials in the fibre solution carrier.²⁰ Understanding more about the evolution from the precursor materials to the electrospun composite fibers can help to forecast appropriate designs, allowing the effective core encapsulation and desired fibre's aspect ratio towards the improvement of their performance in target applications. Previous studies in colloidal electrospinning showed the production of functional surfaces by encapsulating thermoresponsive microgels with different concentrations of particles in the feeding solution.^21,22 Despite of these interesting studies, spinning “smart” microgels within polymers with low or no affinities to each other still presents challenges in design and efficiency of the process.²³ The manipulation of the core-diameter-to-shell-diameter and fibre diameter would allow the production of fibres with the desired aspect ratio.

Therefore, in this work two main objectives are pursued; first, the confinement of thermoresponsive microgels in polymer fibres (poly(ethylene oxide), PEO) in order to obtain composite fibers by means of colloidal electrospinning; and second, we aim to control the architecture of the composite fibers in terms of the aspect ratio by minimizing its diameter using the Response Surface Methodology (RSM) as a statistic tool. With this methodology we will be able to optimize the electrospinning processing parameters²⁴ and to understand their influence on the diameter of the fibres and on the confinement of microgels. Thus, the relationship between the fibre diameter and the process parameters is visualized by response surfaces, which gives access to the relative influence of each variable on the fibre diameter.²⁵

Experimental section

Materials

N-Isopropylacrylamide (NIPAAM, Aldrich Chemistry, 97%) was used as monomer and N,N-methylene bis-acrylamide (MBA, Sigma-Aldrich, 99%) as cross-linker, ammonium persulfate was chosen as initiator (APS, Sigma-Aldrich, 99%) and sodium bisulfite (SBS, Acrös Organics) was used as catalyzer. All reagents were used as received without any further purification. High molecular weight chitosan (CS, 469 kDa, Cognis) was depolymerized using glacial acetic acid (Panreac) and sodium nitrite (NaNO₂, EKA) in order to obtain low molecular weight CS through the oxidative fragmentation.²⁶ Ethanol (C₂H₆O, Scharlau, analytic grade) and poly-ethylene oxide (PEO, Sigma Aldrich, M_w = 5 000 000 Da) were used in the study.

Methods

Colloidal spinning dispersions preparation. PNIPAAM and PNIPAAM–CS cross-linked microgel dispersions were prepared by means of surfactant-free emulsion polymerization (SFEP) in the presence of MBA cross-linking agent following the procedure described in detail elsewhere.²⁷ NIPAAM monomer provides thermosensitivity, while CS prevents the growth of PNIPAAM–CS microgels.²⁸ To prepare PNIPAAM–CS microgels, 50 mL of acetic acid solutions incorporating 1 wt% of chitosan with 30, 50 and 85 kDa were used for different weight ratios of NIPAAM/CS. Table 1 contains in details the recipes for the microgels synthesis.

Table 1 List of the as-prepared microgels and their composition concerning NIPAAM as also the weight percentage and molecular weight of CS

Sample code	NIPAAM (g)	CS^b (wt%)	CS^b (kDa)
a Weight percentages of MBA, APS and SBS were determined as a function of NIPAAM total mass for 10, 10 and 5 wt%, respectively.b CS content was determined in respect to the total mass of NIPAAM using different molecular weights of CS namely 30, 50 and 85 kDa.
PNIPAAM^a	2.5	—	—
PNIPAAM–CS1	1	20	85
PNIPAAM–CS2		30
PNIPAAM–CS3		40	30
PNIPAAM–CS4			50
PNIPAAM–CS5			85

---

Fibre template solution was prepared by dissolving 2 wt% of PEO in the mixture water/ethanol at a ratio of 4 : 1 (v/v). The as-prepared microgel dispersions were mixed to the PEO spinning solution in a ratio of 1 : 1 (v/v) under magnetic stirring during 5 h for further colloidal electrospinning process.

Selected operating conditions for electrospinning experiments. The different as-prepared spinning dispersions were transferred to a 3 mL syringe with a capillary tip fitted with a 23-gauge blunt tip needle and loaded into a syringe pump (KD Scientific, KDS100) programmed to deliver the dispersions within the range of 0.5–0.7 mL h⁻¹. To perform the experiments, a high voltage power supply (Glassman High Voltage, EL, USA) was used in the range of 10–20 kV, whereas the distance between the needle tip and the paper grounded collector was manually adjusted between 13 and 27 cm. The process was carried out at temperatures above of 35 °C which is slightly above the VPTT (volume phase transition temperature) of each microgel dispersion sample under study, whereas the relative humidity was maintained between 20 and 40%. The obtained fibres were vacuum dried at room temperature for 24 h to remove any water residues.

Characterization

The hydrodynamic diameter of the microgels were analysed by dynamic light scattering (DLS) using a Horiba SZ-100 Nanopartica series Analyser Light Scattering instrument equipped with 592 nm He–Ne laser for a scattering angle of 90° and an autocorrelator SZ-100. The swelling properties of microgel dispersions were accessed by increasing the temperature from 25 °C to 42 °C with a Peltier system (25 °C) within disposable cuvette cells. All measurements were carried out for highly diluted dispersions.

The morphology of the as-prepared microgel dispersions as well as the composite electrospun non-woven mats were checked by means of scanning electron microscopy equipped with a Carl Zeiss Auriga CrossBeam system (SEM-FIB). To observe the distribution size of microgels, one drop of aqueous dispersions of each sample under study was deposited on Formvar carbon-coated Cu grid. While the small pieces of the non-woven mats were fixed on conductive carbon tape, mounted on the support and then sputtered with a thin layer of gold/palladium (8–10 nm) using Q300T D sputter coater. For the composite fibres, at least 50 diameters of dried microgels were measured for the minor axis (perpendicular to length fibre) and about 100 fibres were measured to obtain the average diameter per SEM micrograph.

Design of experiments

D-Optimal design. A D-optimal design was performed using JMP version 8.0 (S.A.S. Institute Inc., Cary, NC, USA) to investigate the effect of variables and the interactions between variables on the fibre diameter. The electrospinning experiments were designed through four selected independent variables, namely spinning dispersions, applied voltage, working distance and flow rate. The levels of variables were chosen in the space of parameters that provide suitable and reliable calculations for the optimal distribution of the D-optimal design. Within the array of levels, the software selected a set of experiments on the basis of a full design consisting in a minimum of 27 runs (see ESI†). The overall desirability response was obtained from the calculations of the software within the selected range.

Statistical analysis. The planning and analysis of the colloidal electrospinning experiments were performed within the context of Response Surface Methodology (RSM). All electrospun non-woven mats were morphologically analysed by SEM and then the diameters of the fibres were measured as described early. Results are presented as mean ± standard deviation. JMP version 8.0 was used in data interpretation and graphic image design. The mathematical approximation model that fits the experimental data was determined as described in the ESI.†

Results and discussion

Microgel dispersions as precursor materials

Electrospinning of colloidal systems is a process mainly dictated by the precursor materials composition, stability and homogeneity.^30,31 In this work, since thermoresponsive microgels are the precursor material it is crucial to determine and understand microgels properties to be further electrospun within PEO fibres. In order to test the ability of the microgels to shape the surface of fibres and analyse the influence of their features in the process, the precursor materials were analysed by SEM and DLS techniques.

Morphology analysis. Fig. 1 shows the SEM images of the representative PNIPAAM and PNIPAAM–CS microgels samples containing 20, 30 and 40% of CS, the later one with two different molecular weights (50 and 85 kDa). The spherical shape and homogeneity of most of the synthesised microgels is confirmed. Additionally, the microgels under study presented diameters in the range of 100 nm to 1 μm which fulfil one of the microgel's main features, their colloidal size. A significant decrease in size was observed when comparing PNIPAAM microgels with those containing CS, as presented in Table 2. As expected from previous work,^27,28 the incorporation of such chitosan content derived in a reduction of microgel size to half. Yet, no further conclusions regarding the effect CS molecular weight (M_w) can be drawn since the observed differences are within the experimental error.

Fig. 1 SEM micrographs corresponding to the samples PNIPAAM (A) and PNIPAAM/CS containing 20% (B), 30% (C) and 40% of CS with 50 kDa (D) and 85 kDa (E).

Table 2 Average dried diameter of the as-prepared microgel dispersions with different compositions obtained by SEM and the corresponding volume phase transition temperature (VPTT) measured by DLS

Sample code	SEM diameter (nm)	VPTT (°C)
PNIPAAM	525 ± 20	34
PNIPAAM–CS1	348 ± 45	35
PNIPAAM–CS2	292 ± 59	36
PNIPAAM–CS4	217 ± 75	37
PNIPAAM–CS5	283 ± 55	36

---

Thermosensitive and swelling capacity. DLS was used (Fig. 2) in order to determine the swelling ability of the various microgel dispersions and to examine the effect of the composition in the size and swelling properties. As already stated in previous work²⁷ the incorporation of CS into PNIPAAM microgels does not eliminate their thermosensitivity although highly affects the swelling ability. In Fig. 2A is depicted the evolution of the hydrodynamic diameter with temperature for PNIPAAM and PNIPAAM–CS samples. The microgels diameter decrease with temperature, which is in agreement with the negative responsive behaviour typically demonstrated by this type of microgels. A rise in temperature leads to the deswelling mechanism of the thermoresponsive microgels derived from the free energy competition between the polymer networks and the solvent through the Flory solubility parameter.³² However, the effect of CS in the microgel's hydrodynamic diameter and size is noticeable. For instance, microgels containing CS present a hydrodynamic diameter in the range of 300–500 nm in the swollen state (25 °C) whereas the reference microgel sample, PNIPAAM, shows a hydrodynamic diameter close to 1 μm. Focusing on their swelling behaviour, Fig. 2B describes the evolution with temperature of the relative swelling index (RSI) (d/d₀)³, that is, the average volume of microgel particles at a given temperature over the average volume of the particles in the collapsed state, for the reference sample PNIPAAM and for PNIPAAM–CS samples containing 30 and 40% of CS for 50 and 85 kDa. The intrinsic lower critical solution temperature (LCST) of PNIPAAM polymer (32 °C)²⁸ is still present in the PNIPAAM–CS microgels in which the effect of CS in the swelling properties is noticeable both in the relative swelling and in the abruptness of the response. Volume phase transition of each sample was also collected in Table 2. The incorporation of 30% of CS decreases the microgels' RSI from 14 of the PNIPAM to 5 in the swollen state (25 °C). Furthermore, when incorporating up to 40% of CS the RSI of microgels decreases up to 10 times with respect to PNIPAAM reference sample in the swollen state. This result confirms the already described cross-linker role of CS in this PNIPAAM–CS microgel systems.^27,28 Additionally, the evaluation of RSI allows understanding the effect of CS M_w in the microgel swelling ability. The swelling ability of the microgel sample containing 40% of CS is reduced in half for higher molecular weight (85 kDa) of CS. Similarly to the effect of CS content, we consider that the increase of CS molecular weight could lead to a predominant effect as a cross-linker agent.

Fig. 2 Evolution of the hydrodynamic diameter (d_h) (A) and relative swelling (d/d₀)³ (B) with temperature for PNIPAAM (blue), PNIPAAM–CS2 (grey), PNIPAAM–CS4 (red) and PNIPAAM–CS5 (green) representative samples.

In previous work²⁷ we determined that below 20% of CS no stable dispersions were obtained. Additionally, for the sample containing 20% of CS we determined that there was no sufficient NH₃⁺ groups to ensure the effective complexation of the CS with PNIPAAM, which led to free CS in the system that was detected by an unexpected increase of the hydrodynamic diameter above the VPTT. In the present work a similar unexpected increase in hydrodynamic diameter well above the VPTT was also detected for the sample containing 20% and the one containing 40% of CS with 30 kDa.

To further understand the microgels' behaviour during the electrospinning process we focused in the volume phase transition temperature (VPTT), the critical temperature at which microgel particles start to collapse. Slight differences are observed in the VPTT between PNIPAAM and PNIPAAM–CS samples, in fact the VPTT occurs at 31–32 °C which is the intrinsic LCST of the parent polymer, although a less defined transition is observed as the content of CS is increased.

Electrospun composite fibres

Confinement of microgels in fibers by colloidal electrospinning. One of the biggest hurdles is to optimize the design of electrospun fibres in order to enhance their performance in biomedical applications. As indicated in the Introduction, the case study presented in this paper focused on the fibre confinement of microgels via colloidal electrospinning. We take advantage of the inexpensive and effective electrospinning method to encapsulate microgels in fibers which could provide high aspect ratio and porosity to the final composite membrane and give rise to tuneable and functional surfaces.³³ The use of microgels can also be worthwhile to design the desired topography of the composite fibres. As a proof of concept we used PEO polymer to effectively encapsulate the microgels by colloidal electrospinning, since this is a well-known biocompatible material for solution electrospinning.³⁴ In future work appropriate fibre templates will be used, accordingly to the desired application.

Morphology analysis. Fig. 3 illustrates the SEM images of two of the composite electrospun non-woven mats. The most relevant information obtained is the confirmation of the thermoresponsive microgels confinement within PEO fibres by the observation of beads randomly distributed over the fibres with a typical “bead-on-a-string” morphology.³⁵

Fig. 3 SEM micrograph of PEO fibres incorporating (A) PNIPAAM microgels and; (B) PNIPAAM–CS1 microgels.

Considering that the electrospinning process was performed at temperatures slightly above the VPTT of microgels, the observed diameters of the microgels in the fibres are similar to their hydrodynamic diameter when collapsed (DLS tests in Table 3). This fact could be explained considering that at such temperatures the solvent evaporation in the electrospinning process occurs while microgels are in their collapsed state. The fast evaporation of solvents may “freeze” microgels in such collapsed state which is maintained in the non-woven mat. It is also noticeable an ellipsoidal appearance of the beads often observed in the colloidal electrospinning. This behaviour is a consequence of the electrostatic forces involved in the process, which induced the change of shape of the largest microgels in the Taylor cone.³¹

Table 3 Average diameter of the microgels under study through the DLS measurements as well as the microgels size in the electrospun fibres

Sample code	DLS measurements		SEM measurements
	dh in the swollen state (nm)	dh in the collapsed state (nm)	Microgels' diameter inside the fibers (nm)
PNIPAAM	940 ± 97	396 ± 10	367 ± 100
PNIPAAM–CS1	485 ± 14	329 ± 32	252 ± 105
PNIPAAM–CS2	507 ± 42	286 ± 58	238 ± 70
PNIPAAM–CS3	531 ± 18	363 ± 17	216 ± 54
PNIPAAM–CS4	393 ± 8	300 ± 3	197 ± 54
PNIPAAM–CS5	404 ± 1	292 ± 2	188 ± 52

---

In colloidal electrospinning the fibres' morphologies can be differentiated either as core–shell,³⁶ “bead-on-a-string”,²¹ spindle-like,³⁵ corn-type structures,³⁷ burst-beads³⁸ or a dispersion of the separated phase within the polymer fibre matrix.³⁹ Different factors may induce a fibre morphology change at the selected processing parameters.⁴⁰ In this case, the influence of the composition of the as-prepared microgels can be observed. Comparing the surfaces of the fibres containing PNIPAAM microgels (Fig. 4A and C) to those that incorporate PNIPAAM–CS microgels (Fig. 4B and D) noticeable differences in the roughness of the electrospun non-woven mats are observed. PEO fibres containing PNIPAAM–CS microgels show higher roughness but also reveal the formation of clusters within beads as early presented in SEM images of the corresponding as-prepared microgel dispersions (Fig. 1B–E).

Fig. 4 SEM representative images of the electrospun composite fibres obtained during the DoE with (A) PNIPAAM microgels using 10 kV, 20 cm and 0.5 mL h⁻¹; (B) PNIPAAM–CS5 microgels for 15 kV, 20 cm and 0.6 mL h⁻¹; (C) PNIPAAM microgels for 20 kV, 27 cm and 0.5 mL h⁻¹ and; (D) PNIPAAM–CS2 microgels for 20 kV, 20 cm and 0.7 mL h⁻¹.

Fig. 5A–C shows the evolution of the morphology of the beads in the composite fibres by varying the molecular weight of chitosan, whereas the Fig. 5D–F illustrates the appearance of the beads for microgels with different amounts of chitosan. As a preliminary remark, by increasing both M_w and CS content, the beads evolved from elongated and stretched to protruding clusters with larger volume encompassing several microgels. This might be related to the change from positive to negative surface charge of the PNIPAAM–CS microgels upon addition of chitosan²⁸ which provides favourable conditions to create agglomerates. Nevertheless, the hydrodynamic diameter of the microgels as well as their swelling ability decreased with the incorporation of chitosan, which might favour the production of these agglomerates.

Fig. 5 SEM micrographs of PEO fibres containing (A) PNIPAAM–CS1 microgels; (B) PNIPAAM–CS2 microgels; (C) PNIPAAM–CS5; (D) PNIPAAM–CS3 microgels; (E) PNIPAAM–CS4 microgels and; (F) PNIPAAM–CS5 microgels.

In terms of the optimized electrospun non-woven mat, the fibres do not ensure the effective encapsulation of the microgels as showed by the phase separation between the matrixes and several PNIPAAM cross-linked particles observed in Fig. 6A. Only few of the microgels were kept inside the fibres and the fibres mechanical's quality was compromised. For the systems shown in Fig. 6B and C we observed an average microgel diameter with the twice of the size of the mean fibre diameters. When the jet flows through the needle-tip towards the collector, expanding and bending instabilities take place.³⁵ If the width of the fibre is diminished, the surrounding fluid may no longer have enough mass to fully involve the microgels in its extension, leading to the phase separation with the resulting fibres containing melted interconnections.

Fig. 6 SEM representative images of the PEO electrospun fibres obtained from (A) the optimal window of variables; (B) PNIPAAM–CS3 microgels for 15 kV, 13 cm and 0.7 mL h⁻¹; (C) PNIPAAM microgels for 20 kV, 27 cm and 0.5 mL h⁻¹ and; (D) PNIPAAM–CS4 microgels for 15 kV, 20 cm and 0.7 mL h⁻¹; (E) PNIPAAM–CS3 microgels for 10 kV, 27 cm and 0.7 mL h⁻¹ and; (F) PNIPAAM–CS5 microgels for 20 kV, 13 cm and 0.5 mL h⁻¹.

Fast solvent evaporation can cause morphologic defects such as burst-beads³⁸ and ribbon structures showed in Fig. 6D and E, respectively.^23,41–43 Burst-beads are more frequently noted when a polymeric skin on the jet is formed covering the dispersed particles at first and then collapsing with a quick evaporation of the solvents.³⁸ The sharper spindle-like structures in Fig. 6F were often observed for the PEO fibres incorporating microgels without CS or with the lowest concentration of CS that likewise may decrease the whole polymer concentration.^21,44

Optimization of the electrospun composite fibers. Morphological analysis of the non-woven mats highlighted the difficulties (defects) and numerous parameters that are involved in the development of composite fibers. In fact, the designing of the fibres surface by manipulating not only the precursor materials but also the properties of the fibres as well as the processing parameters still remains unexplored.²³ Further analysis of the processing parameters would provide control over the morphology of the fibres surface towards the manipulation of their properties.²⁴ The incorporation of microgels in fibers, by itself could derive in composite fibres with enhanced aspect ratio if the diameter of the fibre is smaller than the microgel size (nanofiber). For such purpose, a multivariable statistical analysis involving the design of experiments was performed for the PEO fibres confining stimuli-responsive microgels; Response Surface Methodology (RSM) was used to optimize the process since this is a particularly useful approach to explore the effect of the processing parameters taking into account their combined effects or interaction over the evolution of the fibre diameter.^45,46 In this paper RSM was used to optimize and predict the set of variables that will minimize the fibre diameter and thus ensure a high aspect ratio. RSM procedure to optimize the processing parameters for the composite fibres is shown in Scheme 1.

Scheme 1 Diagram of the stages developed in the RSM procedure to optimize the processing parameters of the PEO fibres incorporating microgels without (dark grey) and applying the software JMP (pale grey).

Statistical analysis of the mean fibre diameter (MFD). To evaluate the fibre's diameter and to analyse its own evolution by varying the chosen set of processing variables, a multivariable regression analysis on the basis of the least mean square method was used. Thus, the most suitable experimental model was a first-order model. Quadratic models are better to forecast the value of the aimed response although the examination of mixed (categorical and numerical) variables requires low order polynomials.²⁹ Thereby, after the unknown coefficients (β_s) from eqn (S1)† are estimated, the relation between the MFD algorithm and the input parameters (samples, applied voltage, flow rate and working distance) are described in terms of coded variables is expressed in eqn (S4).†

Statistical reports obtained from the fitting of the model, which are a measure of the amount of variability on the fibre diameter towards the explanatory variables, give rise to the following values: mean 210.3 nm; R² 0.99; R_adj² 0.98; p < 0.0001. The statistical reports R² and adjacent R² indicate how well the model fits the experimental data. The value of R² is a measure of the response being represented by its predictors, which always increases when a new term is added to the model regardless of whether the inclusion of the additional variable is statistically significant or not. On the contrary, the R² report may or not increase when more terms are included, thereby it will increase if the new terms improve the model but will otherwise decrease with insignificant terms. In this work, R² illustrates that 99% of the MFD are explained by the variables whereas the adjacent R² indicates that the model predicts 98% of the variability over the fibre diameter evolution within the selected parameters space.

The accuracy of the model also can be understood by the linear regression of the experimental values of the MFD as a function of the predicted diameters (Fig. 7). From our graphic representation a strong correlation between the model's predictions and its actual results can be observed; fact that supports the good accuracy of the model, as early demonstrated by the almost null values of the model statistical reports. For each modelled effect, an optimal estimative tested by a null hypothesis is given, contributing to the calculations of the equation model. The p-values test the null hypothesis illustrating how statistical significant is a certain estimate which is rejected in a confidence of (1 − p-value) > 0.95 (Table S2† summarizes the main coefficients and their statistical reports).

Fig. 7 Actual versus predicted plot for the mean fibre diameter.

Results from response surfaces for the mean fibre diameter. As mentioned before, RSM allowed exploring the effect of the processing parameters and their interaction over the evolution of the fibre diameter. In this work we explored the individual and interactive effects of spinning solution, applied voltage, working distance and flow rate on the evolution of the fibre's diameter (Fig. 8). In Table 4 the sample code used over the surface responses analysis are represented.

Fig. 8 Response surfaces for the mean fibre diameter in terms of: (A) applied voltage and spinning sample, (B) applied voltage and working distance, (C) working distance and spinning sample, (D) flow rate and working distance, (E) flow rate and spinning sample and (F) applied voltage and flow rate. Design points above () and below () the predictive value.

Table 4 Sample code used over the surface responses analysis in Fig. 8

Variable term	Spinning dispersions code
S0	PNIPAAM microgels/PEO
S1	PNIPAAM–CS1 microgels/PEO
S2	PNIPAAM–CS2 microgels/PEO
S3	PNIPAAM–CS3 microgels/PEO
S4	PNIPAAM–CS4 microgels/PEO
S5	PNIPAAM–CS5 microgels/PEO

---

Influence of the as-prepared microgel samples (spinning sample). Conversion of colloidal systems in composite fibres entails the employment of solutions consisting in two or more immiscible phases. Previous works showed the role of the concentration, homogeneity and size of the as-prepared dispersions in the morphology of the composite fibres.^21,22,31 The spinning systems encompassing microgels of PNIPAAM as well as PNIPAAM containing 30% CS and 40% CS of 85 kDa had influence on the MFD. In the case of PNIPAAM, the highest p-value shows their less significance on the diameter of the fibres. In this case the absence of CS may decrease the whole polymer chain entanglements for which the viscoelastic force is decreased allowing the production of thinner fibres due to the electrostatic stretching forces of the process. At long distances, highly diluted solutions have more time not only to stretch but also to evaporate the solvent favouring the production of thinner fibres. Therefore, the MFD of each sample is vastly dependent on the working distance, as it will be later discussed in this work.

Effect of the applied voltage. As the surface tension of the ejecting polymer droplet is balanced by electrostatic forces, the droplet undergoes a conical deformation and evolves in a polymeric jet towards a grounded collector. While physic laws support that an increase on the applied voltage gives rise to higher stretching forces, experimental findings have shown controversial outcomes by noticing increases, decreases or, even no changes on the fibre diameter as it was described for PEO fibres by Reneker and Chun.^44,47 An increment in the voltage only leads to larger diameters of the PEO fibres confining microgels of PNIPAAM, 30% CS as well as 40% CS of 30 kDa (Fig. 8A). Regarding the interaction between the applied voltage and the working distance (Fig. 8B), it is noticed that combining low distances and high voltages the MFD increases, which is in agreement with the presence of the Vd term in the MFD algorithm.

Effect of working distance. The working distance is related to the solvent evaporation and fibre formation. When the jet travels towards the collector it is subjected to strong thinning and whipping phenomena due to solvents evaporation and electrostatic repulsive forces.²³ A slight increase of the distance between the tip and collector will promote a decrease in the diameters, although it is vastly dependent on the polymer concentration.⁴⁴ In this study, even when separated from the other variables, the working distance influenced the MFD. The effect of working distance affects the MFD of the PEO fibres with microgels of PNIPAAM and 40% CS of 85 kDa, as it is showed in Fig. 8C proving the existence of dS₀ and dS₅ respectively.

Influence of the flow rate. To start the process it is necessary to have a minimum flow to obtain a droplet of the solution at the tip of the needle in order to form a stable Taylor cone.⁴² From our computational modelling it is observed that the flow rate can induce changes on the MFD. Regarding the interaction between the working distance and the flow rate, the fibre diameter undergoes two effects: (i) combining low flow rates and short distances derives in the production of smaller diameters; (ii) at high flow rates and low distances thinner fibres were obtained. Looking at the response surfaces in Fig. 8D–F, there is a huge interaction between the flow rate and the PNIPAAM microgel sample as well as PNIPAAM microgels samples containing 40% CS (30 and 85 kDa), a slight interaction between the flow rate and the PNIPAAM microgel dispersions with 30% CS, and no interaction between the flow rate and the voltage, as it was already showed by the presence of VS₀, VS₂, VS₃, VS₄, VS₅ and absence of VQ terms.

Optimization of the fibre diameter. The overall settings led to the production of PEO fibres with diameters ranging from 136 ± 72 nm to 309 ± 107 nm with the set of variables [PNIPAAM–CS1 microgels/PEO, 10 kV, 20 cm and 0.7 mL h⁻¹] and [PNIPAAM microgels/PEO, 10 kV, 20 cm, 0.5 mL h⁻¹], respectively. In the basis of the analysis of the operational conditions and their effect on the MFD the software also conducted calculations to estimate the set of variables that warrants the minimization of the diameter of the fibres. The goal response was predicted for the following settings: 15 V, 20 cm, 0.6 mL h⁻¹ using the spinning sample of PNIPAAM microgels/PEO. By carrying out the process with the corresponding values, ultra-fine fibres with a MFD of 63 ± 25 nm were obtained, as it is observed and calculated from the image shown in Fig. 9.

Fig. 9 SEM image optimized electrospun fibres containing PNIPAAM microgels (on the right) and the fibre diameter distribution with a MFD of 63 ± 25 nm (on the left).

Conclusions

This work demonstrates the effective confinement of thermoresponsive microgels in PEO electrospun fibers via colloidal electrospinning. The statistical analysis provides an effective means to study and predict the behaviour of the process towards the production of ultra-fine composite fibres. Using RSM the number of experiments was significantly reduced, generating a larger amount of information from a small number of experiments while maintaining a high level of analysis accuracy. The test of the forecasted variables aimed to minimize the size of the fibres, which resulted in the production of PEO nanofibres with a mean diameter of 63 nm, proving the ability of the procedure to predict the optimum conditions. The reduction of the fibre size to less than half of the microgel's diameter may compromise the effective encapsulation of such dispersed particles in the fibre carrier solution. Further experiments show that increasing both M_w and the content of CS in the precursor microgel dispersions affected their distribution within the PEO fibre changed from randomly distributed to agglomerates. The overall study on the morphology showed outstanding results to tailor the architecture of the composite fibres when compared with the state-of-art without statistical approaches. This report opens new strategies to design desired architectures via colloidal electrospinning by taking advantage of thermoresponsive microgels as potential active sites in composite electrospun fibers with potential biomedical applications such as drug delivery systems, biosensors or even self-healing materials.

Acknowledgements

This work is funded by FEDER funds through the COMPETE 2020 Programme and National Funds through FCT – Portuguese Foundation for Science and Technology under the project UID/CTM/50025/2013. We also acknowledge support from the Portuguese Science and Technology Foundation through grant SFRH/BPD/88779/2012 (CE).

Notes and references

1. B. R. Saunders and B. Vincent, Adv. Colloid Interface Sci., 1999, 80, 1–25.
2. B. R. Saunders, N. Laajam, E. Daly, S. Teow, X. Hu and R. Stepto, Adv. Colloid Interface Sci., 2009, 147–148, 251–262.
3. A. Pich and W. Richtering, in Advances in Polymer Science, 2010, vol. 234, pp. 1–37.
4. R. Pelton, Adv. Colloid Interface Sci., 2000, 85, 1–33.
5. A. Imaz and J. Forcada, J. Polym. Sci., Part A: Polym. Chem., 2008, 46, 2510–2524.
6. J. Ramos, A. Imaz, J. Callejas-Fernández, L. Barbosa-Barros, J. Estelrich, M. Quesada-Pérez and J. Forcada, Soft Matter, 2011, 7, 5067–5082.
7. C. Echeverria, D. López and C. Mijangos, Macromolecules, 2009, 42, 9118–9123.
8. M. Bradley, J. Ramos and B. Vincent, Langmuir, 2005, 21, 1209–1215.
9. C. Johansson, J. Gernandt, M. Bradley, B. Vincent and P. Hansson, J. Colloid Interface Sci., 2010, 347, 241–251.
10. C. Echeverria and C. Mijangos, Macromol. Rapid Commun., 2010, 31, 54–58.
11. C. Echeverria and C. Mijangos, Langmuir, 2011, 27, 8027–8035.
12. C. Echeverria, P. Soares, A. Robalo, L. Pereira, C. M. M. Novo, I. Ferreira and J. P. Borges, Mater. Des., 2015, 86, 745–751.
13. B. Mizrahi, S. Irusta, M. McKenna, C. Stefanescu, L. Yedidsion, M. Myint, R. Langer and D. S. Kohane, Adv. Mater., 2011, 23, H258–H262.
14. J. K. Oh, R. Drumright, D. J. Siegwart and K. Matyjaszewski, Prog. Polym. Sci., 2008, 33, 448–477.
15. W. Wu, J. Shen, Y. Li, H. Zhu, P. Banerjee and S. Zhou, Biomaterials, 2012, 33, 7115–7125.
16. W.-F. Lai, A. S. Susha and A. L. Rogach, ACS Appl. Mater. Interfaces, 2016, 8, 871–880.
17. A. B. South and L. A. Lyon, Angew. Chem., Int. Ed., 2010, 49, 767–771.
18. J. C. Gaulding, M. W. Spears and L. A. Lyon, Polym. Chem., 2013, 4, 4890–4896.
19. M. W. Spears, E. S. Herman, J. C. Gaulding and L. A. Lyon, Langmuir, 2013, 30, 6314–6323.
20. C. Wang, L. Wang and M. Wang, Mater. Lett., 2014, 124, 192–196.
21. S. Y. Gu, J. B. Li, J. Ren and Z. M. Wang, Macromol. Mater. Eng., 2010, 295, 32–36.
22. J. E. Diaz, A. Barrero, M. Marquez, A. Fernandez-Nieves and I. G. Loscertales, Macromol. Rapid Commun., 2010, 31, 183–189.
23. D. Crespy, K. Friedemann and A.-m. Popa, Macromol. Rapid Commun., 2012, 33, 1978–1995.
24. O. S. Yördem, M. Papila and Y. Z. Menceloğlu, Mater. Des., 2008, 29, 34–44.
25. M. Mohammadian and A. K. Haughi, Bulg. Chem. Commun., 2014, 46, 545–555.
26. P. I. P. Soares, A. I. Sousa, J. C. Silva, I. M. M. Ferreira, C. M. M. Novo and J. P. Borges, Carbohydr. Polym., 2016, 147, 304–312.
27. C. Echeverria, P. Soares, A. Robalo, C. M. M. Novo, J. P. Borges, I. Ferreira and L. Pereira, Mater. Des., 2015, 86, 745–751.
28. M. K. Jaiswal, R. Banerjee, P. Pradhan and D. Bahadur, Colloids Surf., B, 2010, 81, 185–194.
29. M. A. Bezerra, L. a. Escaleira, E. P. Oliveira, R. E. Santelli and L. S. Villar, Talanta, 2008, 76, 965–977.
30. M. A. Badawi and L. K. El-Khordagui, Eur. J. Pharm. Sci., 2014, 58, 44–54.
31. M. Angeles, H. L. Cheng and S. S. Velankar, Polym. Adv. Technol., 2008, 19, 728–733.
32. J. Wu and Z. Hu, Dekker Encyclopedia of Nanoscience and Nanotechonology, 2004, pp. 1967–1976.
33. J. Wu, N. Wang, Y. Zhao and L. Jiang, J. Mater. Chem. A, 2013, 1, 7290.
34. X. Hu, S. Liu, G. Zhou, Y. Huang, Z. Xie and X. Jing, J. Controlled Release, 2014, 185, 12–21.
35. H. Qi, P. Hu, J. Xu and A. Wang, Biomacromolecules, 2006, 7, 2327–2330.
36. A. V. Bazilevsky, A. L. Yarin and C. M. Megaridis, Langmuir, 2007, 2311–2314.
37. S. Agarwal and A. Greiner, Polym. Adv. Technol., 2011, 22, 372–378.
38. S. Maretschek, A. Greiner and T. Kissel, J. Controlled Release, 2008, 127, 180–187.
39. S. Yan, L. Xiaoqiang, L. Shuiping, M. Xiumei and S. Ramakrishna, Colloids Surf., B, 2009, 73, 376–381.
40. M. P. Prabhakaran, S. Ramakrishna and M. Zamani, Int. J. Nanomed., 2013, 8, 2997–3017.
41. F. Elahi, G. Guoping, F. Khan and W. Lu, J. Bioeng. Biomed. Sci., 2013, 3, 1–14.
42. Y. E. Choonara, C. Dott, V. M. K. Ndesendo, L. C. du Toit, V. Pillay, L. Tomar, P. Kumar and C. Tyagi, J. Nanomater., 2013, 1–23.
43. D. N. Rockwood, D. B. Chase, R. E. Akins and J. F. Rabolt, Polymer, 2008, 49, 4025–4032.
44. N. Bhardwaj and S. C. Kundu, Biotechnol. Adv., 2010, 28, 325–347.
45. H. M. Khanlou, A. Sadollah, B. C. Ang, J. H. Kim, S. Talebian and A. Ghadimi, Neural Computing and Applications, 2014, 25, 767–777.
46. S. Sukigara, M. Gandhi, J. Ayutsede, M. Micklus and F. Ko, Polymer, 2004, 45, 3701–3708.
47. I. Chun, H. Fong and D. H. Reneker, Polymer, 1999, 40, 4585–4592.

---

Footnote

† Electronic supplementary information (ESI) available: The defined design of experiments. See DOI: 10.1039/c6ra12713d

---