Stroboscopic microscopy—direct imaging of structure development and phase separation during spin-coating



Spin-coated polymer blends possess a rich variety of accessible non-equilibrium morphologies, formed through a process of phase separation and self-assembly, the complexities of which remain incompletely understood. The technique of stroboscopic microscopy has now been developed to allow direct observations of microscopic and mesoscopic morphological development during spin-coating and has afforded unequivocal information regarding morphological development. The technique so far has three modes of operation providing information on topographical, compositional, and crystal development. In this review, we look at the technique's development, its applications and comment on the future potential for this technique. © 2013 Wiley Periodicals, Inc. J. Polym. Sci., Part A: Polym. Chem. 2014, 52, 17–25


Spin-coating is a facile technique for the production of high-quality, uniform thin-films. As such, the technique finds a wide range of applications including semiconductor wafer production (photoresists),[1] organic electronics,[2-6] optical coatings,[7-9] and in biosensors[10] to name but a few. Spin-coating involves depositing a polymer solution onto a substrate that is subsequently rotated at high speed (typically between 1000 and 10,000 rpm). Initially, film thinning occurs due to rotational forces[11] (hydrodynamic thinning) and solution is ejected from the edge of the rotating surface. At some point, due to an increasing viscosity, an equilibrium is reached where hydrodynamic thinning is minimal and film thinning predominantly occurs through evaporation (evaporative thinning).[12] A number of parameters affect the interplay between hydrodynamic and evaporative thinning including viscosity, vapor pressure,[13] molecular weight, acceleration rate,[14] and the rotational rate.[15, 16]

Spin-coating a binary mixture (e.g., Polymer A + solvent) usually results in the formation of a uniform film with a continuous morphology. Many early in situ studies of this process focused on understanding film thinning behavior of pure solvent and binary mixtures.[17-19] When ternary mixtures (Polymer A + Polymer B + Solvent) are spin-coated, often complex processes of self-assembly take place, such as phase-separation,[20-22] crystallization,[23] stratification,[13, 24-26] and agglomeration, resulting in the formation of intricate morphologies that evolve toward thermodynamic equilibrium, but which ultimately become frozen due to rapidly increasing viscosities.[13, 27] The propensity of polymer blends to self-assemble, forming well-ordered morphologies is highly desirable, particularly for organic electronic devices, such as organic photovoltaics, where a morphology with a large amount of interface between phases and interconnectivity is required for the generation and separation of charges.[5] The pursuit of producing such ordered morphologies has driven research aimed at furthering our understanding of self-assembly, so that processing conditions may be designed rationally, allowing for a greater degree of control over device morphology. It should be noted that a great deal of work has been undertaken to look at phase transitions in polymer melts.[28, 29] However, such systems are two component systems (Polymer A and Polymer B) and so do not possess the high levels of mobility as when a solvent is present. Moreover, the majority of polymer based electronic devices are solvent processed, as opposed to melt processed, and hence, the focus of these experiments is to look to replicate real world production methods and use these findings to better understand and direct such processes.

The final morphology obtained from a spin-coated polymer blend is often evaluated postprocessing and is used to elucidate, via an informed hypothesis, how morphological development proceeded. Owing to the high speed and non-equilibrium nature of spin-coating, this approach is far from ideal. Studies conducted in situ have been crucial in furthering our understanding of the processes that take place during spin-coating[13, 16, 26, 30-32] and we have summarized these findings in a recent review.[33] The majority of in situ studies have been based upon laser interferometry and scattering, affording information regarding film thinning behavior and morphological development, respectively. However, a fundamental drawback of techniques based upon scattering is that results are based in reciprocal space, and as such, information regarding morphological development must be inferred from changes in the observed scattering length scale.

In order to overcome this problem and obtain results based in real space, a stroboscopic microscopy technique (dubbed the “optostrobometer”) was developed to allow direct observations of the complex processes of self-assembly, taking place during spin-coating. The technique works by applying a short light pulse (∼50 μs) while imaging near the center of rotation, once per complete revolution. (At 1500 rpm, a light pulse of 50 μs will result in 0.785 μm of movement and 100 μm from the center of rotation.) As such, a static image is observed, despite the high rotation rate. The technique has provided unequivocal information on how such a rich variety of morphologies may develop, from what may appear to be a relatively simple processing route. This article charts the technological developments of the optostrobometer technique for the study of morphological development of polymer blends.

To our knowledge, the first example of a stroboscopic approach being applied for visualization during spin-coating was performed by Graves and co-workers, where a laser and a 35-mm film camera was synchronously pulsed, producing a series of interferometric photographs.[34, 35] These studies provided insight into spin-coating over complex topographies and were motivated by the fact that previous processing steps in semiconductor device manufacture, often left uneven surface topographies. However, this technique has only recently become practical due to the emergence of high power LEDs and high quantum efficiency electron multiplying charged coupled device (EMCCD) cameras, disposing the need for high-powered lasers and photographic film.

Stroboscopic microscopy utilizing LED illumination was first developed by the Howse research group.[31] Owing to the wide variety of commercial LEDs available, so far, it is now possible to extend the technique for the study of topographies,[16, 31, 36] composition,[32] and crystallization.[37] These modes of operation, along with representative experimental data are summarized in Table 1.

Table 1. Optostrobometer Modes of Operation and Representative Data
Mode of OperationLEDFilterData
TopographyNear monochromatic red/greenBand-pass filter ∼20 nm, centered on LED λmax
CompositionUV-blueBeam cube comprising of appropriate filters for fluorophore to be studied
CrystallizationWhite (high flux due to effect of polarizers)Polarizer with second analyzer polarizer


The technique utilizes a standard episcopic microscope, with a DC motor mounted directly underneath the microscope objective, a high-powered LED, and a high-sensitivity EMCCD camera that are both electronically synchronized with the motor's rotation. This setup along with the signal processing flow structure is represented in Figure 1.

Figure 1.

Schematic of the stroboscopic microscope setup.

Synchronization between the spin-coater, LED, and camera is achieved by monitoring periodic oscillations in the motor current arising due to the commutator action. A series resistor converts the current to a voltage waveform (Fig. 1-i). A waveform of six cycles per revolution is typical of motors used to date. The voltage waveform displays sharp dips at a key point per cycle as the motor brushes pass between commutator contacts. A hardwired inverting differentiator (Fig. 1-ii) and inverting threshold comparator detect these transitions, producing a pulse train (Fig. 1-iii). Spurious peaks are removed by using these pulses to trigger a fixed-duration pulse generator. The resulting signal (Fig. 1-iv) is used as a tachometer and through incorporation of feedback stabilizes the motor RPM. The cleaned pulse train is divided by 6 via a decade counter (Fig. 1-v) and square-wave pulses of a predefined length (dictated by the desired illumination pulse time) are generated from it (Fig. 1-vi). These pulses are subsequently amplified to drive the illumination LED and trigger the camera. This approach allows the use of any DC motor to rotate the substrate. The LED and camera are typically pulsed for 50 μs, depending upon experimental conditions (rotation rate, objective employed); such an illumination time is sufficient to capture an image with minimal motion blur at 1500 rpm. This is achieved with repeatable positional accuracy below the optical limit, once per rotation. By selecting appropriate LEDs and optical filters, we have developed three modes of operation, for the study of topographies, composition, and crystallization.

Topographical Imaging

When the setup incorporates a near monochromatic LED, with a narrow bandpass filter, contrast in the obtained images reflects difference in optical thickness (nd) where n is refractive index and d is film thickness. For polymers with similar refractive indicies, the contrast within the image therefore results from topographical height differences alone. Through postprocessing, it is possible to generate drying curves (film thickness as function of time) for each pixel and therefore it is possible to generate, three-dimensional topographies [shown in Fig. 2(c)] as a function of time. Such curves/topographies are generated through analysis of the variation in pixel intensity, which oscillates through consecutive constructive maxima/minima as the film thins. Constructive interference is defined by a simplified form of the Bragg equation, d=mλ/2n, where d is the optical thickness, m is an integer value, λ is the incident wavelength, and n is the refractive index. The sensitivity of this interferometric technique is dependent on the wavelength and monochromaticity of the illumination source. The effect of the wavelength distribution is that the magnitude of the intensity between constructive and destructive interference increases as the film thins, giving rise to an oscillating amplitude, as defined by:[38]

display math

where I is the intensity and Δk is the full width of the wavenumber (k = 2π/λ).

Figure 2.

(a) Series of unprocessed stroboscopic optical reflectance images recorded during spin-coating of a 1:1 blend of PS:PI (2% wt at 1500 rpm) from o-xylene. Corresponding background-corrected two-dimensional Fourier transforms are shown underneath each frame. (b) Radial integral of the Fourier transform for the entire movie. Taken and adapted from ref. 31.

As such, an LED with a small wavelength distribution will result in higher amplitude oscillation in intensity while longer wavelength radiation results in consecutive maxima being spaced further apart (in time). Such a combined approach allows features to be more easily observed at larger film thicknesses and film thinning rates as encountered with high vapor pressure solvents. Conversely, illumination with a broad wavelength distribution LED, would produce a small amplitude signal with a low signal-to-noise ratio at large film thicknesses offering only a limited insight into the process.

The interferometry-based setup has so far been exploited for the study of the model blends of polystyrene (PS) with polyisoprene (PI) and PS with poly(methyl methacrylate) (PMMA) as shown in Figures 2 and 3, respectively.[16, 31]

Figure 3.

(a) Direct observation of phase separation in blends of PS:PMMA (1:4 by weight) spun-cast (3 wt % at 1500 rpm) from o-xylene, (i–xii) 0.00, 5.04, 5.48, 6.00, 6.52, 7.04, 7.60, 8.16, 8.72, 9.84, 12.56, and 15.60 s, respectively; and (b) corresponding radial integrals of Fourier transformed data as a function of time showing the morphological development. Taken from ref. 16.

For the blend of PS:PI, a bicontinuous morphology, indicative of spinodal decomposition, was formed relatively early on in the drying process at a film thickness of approximately 2 μm [Fig. 2(a)], around 10 times the final film thickness. The bicontinuous structure showed little evolution once first formed, with the height reconstruction data corroborating an earlier hypothesis that phase separation originates at the surface and that this initial structure acts as a motif, directing subsequent ordering of the underlying regions.

The early phase separation behavior of the PS:PMMA blend was similar to the PS:PI blend, forming an initial bicontinuous morphology, when the film was relatively solvent rich [Fig. 3(a-ii–iv)]. However, a high degree of domain growth and coalescence subsequently occurred [apparent by the faint “whisker” in the radially integrated FFT data shown in Fig. 3(b)], resulting in the formation of a bicontinuous morphology with a much larger length scale than first formed through spinodal decomposition [Fig. 3(a-v–xii,b)]. The alternative behavior of the PS:PI and PS:PMMA blends was thought to be due to the differences in surface free energies of PS, PI, and PMMA. For the blend of PS:PMMA, the difference in the surface free energies between the respective homopolymers is greater than that of the PS:PI blend. For the PS:PMMA blend, there is therefore a larger driving force for domain coarsening, as the system moves to a more equilibrated morphology, minimizing the unfavorable interactions between polymers. Conversely, in the PS:PI blend, this driving force is much weaker, leading to a morphology that does not alter significantly from its formation.

In addition, the PS:PMMA blend was studied at a range of rotation rates (1500–5500 rpm), with results indicating that increasing the rate of rotation led to a higher rate of evaporation. Consequently, this forced a deeper quench through the phase diagram, acting to freeze morphological development earlier on, thus producing morphologies further from their thermodynamic equilibrium.

Crystallization Imaging

Polymer chains often crystallize through either π−π stacking or the formation of chain folded lamellae that may additionally self-organize to form high order structures known as spherulites, which grow out from central nuclei. Spherulites are highly birefringent due to the high degree of ordering and can be studied through use of polarized white light, with a second polarizing analyzer. Through the implementation of a high-intensity white LED with crossed-polarizers, it was possible to directly visualize spherulitic crystal growth during spin-coating, providing information regarding crystallization kinetics during processing. Such work is motivated by applications in organic electronic devices, where device efficiency is linked to the formation of a well-defined nanoscale phase-separated morphology with a high degree of crystallinity, facilitating the movement of charges.[5] The crystallization of a model polymer system, poly(ethylene glycol) (PEG), was studied through incorporating polarized white light illumination with a second polarizing analyzer. The rate of crystallization was obtained by measuring the change in the spherulite radius.[37] Results showed that higher polymer concentrations led to a delay in the onset of crystallization yet the rate of crystallization remained constant, consistent with isothermal crystallization. It was believed that the spontaneous development of crystal nuclei only occurred when the concentration became sufficiently high.

The interplay between phase separation and crystallization was studied in blends of PEG:PS, spun-cast from chloroform, by employing tandem experimental runs of normal monochromatic interferometry (phase separation) with crossed polarizers (crystallization). Polymer blends at compositions of 7:3, 1:1, and 3:7 (PEG:PS) and different molecular weights of PEG(4k):PS(28k) and PEG(10k):PS(28k) were studied, and showed a rich variety of morphological development. Results for the PEG(10k):PS(28k) at compositions of 7:3, 1:1 are shown in Figure 4.

Figure 4.

Direct observation of phase separation and crystallization in blends of 10k PEG: 28k PS spun-cast from 8 wt % chloroform solution at 1500 rpm with a ramp of 10 s at compositions of (a) 7:3 and (b) 1:1. For each, showing interferometric (top) and crossed polarized micrographs (bottom). Taken from Ref. 37.

For the 4k 1:1 blend, morphological development was characterized by the onset of a bicontinuous interconnected morphology, with subsequent growth and coarsening. Crystallization occurred after the final phase separated morphology became fixed. The 3:7 blend with 4k PEG formed a more disordered morphology with no common length scale and formed dendritic-type crystals. Both the 4k and 10k 7:3 [Fig. 4(a)] blends initially formed hexagonal/pentagonal cells, known as Bérnard cells, characteristic of Marangoni phenomena. Phase separation proceeded within the Bérnard cells, via a nucleation and growth mechanism, resulting in the formation of large discrete islands of one phase within a continuous matrix of the other. The 10k 1:1 [Fig. 4(b)] and 3:7 blends formed Bérnard cells that were not interconnected and lay within a matrix of polymers phase separating through spinodal decomposition. Phase separation within the Bérnard cells occurred later than outside of the cells. However, domain growth progressed rapidly within the Bérnard cells once phase separation had been triggered. The final morphology was characterized by nucleated domains of one phase with two differing length scales, in a continuous matrix of the second phase.

Composition Imaging

One of the drawbacks with the previously mentioned interferometry approach is that constructive interference is dependent on optical thickness (n, d), where d is the thickness and n is the refractive index. As such, observed structural developments cannot be exclusively attributed to either topographical or compositional fluctuations when there is more than an insignificant difference between refractive indicies of the two polymers. Through studying the observed fluorescence of a fluorescent polymer in a blend comprising of fluorescent + nonfluorescent polymers, the observed contrast reflects composition, exclusively. The fluorescence setup, requires an LED with an emission wavelength matching the excitation wavelength of the fluorophore studied, with an appropriate filter-cube (excitation filter, dichroic mirror, and emission filter) to effectively separate the excitation from the emitted wavelengths.

This fluorescence technique was first used to study morphological development in the blend of poly(9,9′-dioctylfluorene) (PFO) and PS, spun-cast from o-xylene.[32] PFO and related co-polymers have been investigated for applications in light emitting diodes and organic photovoltaic devices.[3, 39, 40] An LED with an emission of 360 nm was used to excite PFO and a Nikon UV-2A beamcube was employed to separate the emitted fluorescence. PFO:PS blends were studied at compositions of 55:45, 60:40, and 65:35 and were spun-cast on silicon and glass. At all compositions an initial bicontinuous morphology formed through spinodal decomposition, which subsequently grew and coalesced through Ostwald ripening [as shown in Fig. 5(a,b)] for the 55:45 and 60:40 blends, respectively). At a critical point during the spin-coating, an instability led to the break-up of the bicontinuous morphology, resulting in the formation of discrete islands within a continuous matrix (islands of PFO, matrix of PS for 55:65 and 60:40 blends, the inverse case for the 65:35 blend). The islands subsequently then grew through Ostwald ripening. For the 60:40 blend, a significant degree of coalescence allowed a bicontinuous morphology to be reformed. In the 55:35 and 65:35 blends, interconnectivity was not re-established, with the final morphologies characterized by discrete large nucleated islands or elongated islands that are not fully interconnected, respectively. The instability showed a compositional dependence occurring earlier for blends containing high PS contents. It was believed that when the instability occurred earlier, when the viscosity was lower, a fast rate of diffusion triggered the formation of submicron islands. Conversely, when the instability occurred later (when the viscosity was higher), the rate of diffusion was slower, resulting in domains “shrinking” away from the bicontinuous morphology. A final interconnected morphology was observed for the 60:40 blend as the instability occurred sufficiently early to allow a large degree of phase coarsening to take place without producing structures that were smaller than the initial spinodal decomposition.

Figure 5.

Direct observation of phase separation in blends of PFO:PS spun-cast (2 wt % at 1500 rpm) from o-xylene at compositions of (a) 55:45 and (b) 60:40; stroboscopic fluorescence images (i–vii) at 8.04, 8.28, 8.72, 9.20, 10, and 15 s, respectively; static (vii) fluorescence images and (viii) AFM of final morphologies, (ix) radially integrated Fourier transform for a selection of each video, vertical lines correspond to times for images (i–vi). (c) Direct observation of phase separation in blends of PFO:PS (65:35) (2 wt % at 1500 rpm) from o-xylene with (circle) and without (square) feedback. Stroboscopic fluorescence images at 10.60, 11.60, 11.80, 12.40, 13.12, 13.75, and 15.00 s. (d) Radial speed in RPM as a function of time with (dotted) and without (solid) feedback. (e) Final (20×) fluorescence images. Taken and adapted from ref. 32.

Additionally, the technique was developed to employ real-time fast Fourier transform analysis of the acquired images, allowing the length scale of phase separation to be tracked during spin-coating. This facilitated the implementation of in situ feedback, where the quench through the phase diagram was altered, through control of rotation rate, in order to obtain targeted morphologies. This approach for obtaining targeted morphologies was demonstrated for the 65:35 PFO:PS blend, which formed a morphology which did not exhibit a high degree of interconnectivity. Figure 5(c,d) shows in situ stroboscopic fluorescence data and the spin profiles for a 65:35 blend spun cast with [Fig. 5(c), circle], and without [Fig. 5(c), square], feedback. It shows that both blends behave similarly, forming bicontinuous morphologies that grow and coarsen. At 11.6 s, feedback increases the rotation rate to 3000 rpm, and at 12.4 s, there is a discernible difference between the blend morphologies. The final ×20 fluorescence image [Fig. 5(e)] shows that the 65:35 blend spun with feedback has a much greater degree of interconnectivity than the same blend spun at 1500 rpm.


Such in situ studies, as described here, have been vital in developing our knowledge of the many parameters that determine how morphological development proceeds during spin-coating. The optostrobometer and its various modes of operation have offered unequivocal information on the spin-coating process in a wide range of polymer systems and have been developed to operate in topographical, compositional and crystallization modes. As with the development of microscopy itself, it is proving to be as versatile as standard microscopy, with various combinations of filters, wavelengths, and polarizers, revealing various information about the structures forming during spin-coating. However, in spite of recent progress we still do not fully understand all of the nonequilibrium processes that take place. There remains, therefore, a large amount of work to be done until we can fully understand and control the processes, which occur during spin-coating.


  • Image of creator

    Daniel Toolan studied chemistry and completed his Master's thesis under the supervision of Tony Ryan at the University of Sheffield in 2011. He is currently in the final year of his PhD, which aims to understand and control morphological development in spin-coated polymer blends and has additionally been working developing various in situ measurement techniques for the study of a number of soft condensed matter systems.

  • Image of creator

    Richard Hodgkinson completed his Chemical Engineering Master's thesis in 2012 at The University of Sheffield under the guidance of Dr Jonathan Howse. During the degree and the associated thesis project, hardware was developed for spin-coating visualization techniques. He is currently in the first year of his PhD, investigating rheometry instrumentation approaches, with an emphasis on techniques involving cross-discipline contribution.

  • Image of creator

    Jonathan Howse obtained his PhD in 2000 (Physical Chemistry, Sheffield). Following a 2-year research position at the Berlin Neutron Scattering Centre and the TU-Berlin, he returned to Sheffield to conduct research on a variety of “soft-nanotechnology” projects. He was appointed lecturer in 2007 and has since led pioneering work on stroboscopic microscopy for looking at spin-coating (discussed here), polymer vesicle formation, and colloidal nanoswimmers.