Threshold response using modulated continuous wave illumination for multilayer 3D optical data storage

Commercial optical data storage (ODS) technologies have found a widespread application for personal data storage and multimedia distribution. However, their application for enterprise data storage has been limited by the storage capacity, arising from limited access to the axial dimension due to a small number of data layers (less than 4) and the limitations in areal density imposed by the diffraction limit. There have been efforts to increase the storage capacity, including holographic storage,^1,2 spatial³ and polarization multiplexing,⁴ and DNA-based storage.⁵ Another widely studied approach is to increase access to the axial dimension by involving nonlinear optical processes such as two-photon absorption and reverse saturable absorption, which confine the writing of bits only to one layer. These approaches have demonstrated the technical viability of many layer storage,^6,7 however, these nonlinear responses are weak and typically require high power pulsed lasers that increase the cost, size, and have safety drawbacks. Thus, a viable many layer ODS approach has been difficult to commercialize.

To develop an efficient 3D data storage medium, it is helpful to consider the operating principles of existing technology. Current writable commercial systems rely on the modulated reflection patterns from metal films due to an adjacent phase change material, usually an inorganic dielectric or organic dye.^8,9 Absorption of the light causes localized heating, which results in a change in phase of the material (typically crystalline to amorphous) which then alters the reflection from the adjacent metal layer.¹⁰ Disc capacity can be increased by stacking more than one data layer, but this approach is difficult because of high attenuation per layer in the reflection-based data scheme¹¹ as well as the high fabrication complexity¹² and resulting costs.

We have previously reported on a method to overcome these limitations by showing that a co-extruded multilayer polymer film can be used as a 3D data storage medium. Data are written by attenuating the fluorescence (FL) intensity of a fluorophore incorporated in thin active layers separated by thicker buffer layers.¹³ In this case, the nanoscopically thin active layers allowed by the co-extrusion process resulted in sufficiently low absorption so that the propagating beam could access 23 layers. Thus, this writing method is suitable for incorporation into the co-extrusion process, in which dozens of writable layers can be fabricated at low cost, provided the materials are relatively simple.¹⁴ 

In this previous work, the writing process was seen to change character from photochemical bleaching on long time scales to a different, and nonlinear process, on shorter time scales. The work reported here aims to investigate the physical nature of this nonlinear writing process by using thin, spun coat, fluorophore-doped polymer films. Thin films are used as opposed to multilayer films so that they are more accessible to a broad range of analytical techniques. The results will be used to inform later studies on co-extruded films where the written data is buried inside the polymer.

The writing process is similar to commercial discs in that it uses a thermal process, but bits are based on FL modulation which, as opposed to reflection, is more suitable for accessing many layers. The FL of the film is changed using a photothermally mediated permanent FL attenuation (FLA) of the fluorophore, which as we describe here, exhibits a threshold dependence on the laser intensity. Data are written within the linear absorption band of the dye using a modulated continuous wave (CW) Blu-ray™-like laser. A threshold process is of great importance in 3D ODS as it localizes the response to the focal volume eliminating interlayer crosstalk and thus allows many layers to be written and read more often. A threshold response also allows writing below the diffraction limit leading to high areal data density.¹⁵ Furthermore, the process is largely independent of the type of dye molecule, provided a sufficient temperature can be achieved, allowing for a large degree of freedom in designing an ODS medium.

Three different dyes were chosen: Rhodamine 6G (R6G), Perylene Orange (PO), and 4-(4-(4-hydroxypiperidin-1-yl)styryl)-2-(dicyanomethylene)-2,5-dihydro-5,5-dimethylfuran-3-carbonitrile (TK01093). R6G was chosen because it is well studied,¹⁶ and PO because it is relatively photostable.¹⁷ Since these dyes do not have a strong absorption at 405 nm, a modulated 488 nm CW laser was used to provide a large enough single photon absorption in the films. TK01093 is a custom dye that has a broader absorption spectrum. We have previously examined the dyes similar to TK01093 containing the dicyanomethylenedihydrofuran acceptor for use in photorefractive media¹⁸ and in single molecule FL studies.¹⁹ The molecular structures and weights of the dyes used in the current study are shown in Fig. 1. Details on the extinction and FL properties of these dyes can be found in the supplementary material.

Films containing poly(styrene-co-acrylonitrile) (SAN25) doped with the fluorescent dyes R6G, PO, and TK01093 at different molar concentrations were used for photoinduced FLA experiments. R6G was obtained from Sigma Aldrich and PO was obtained from Kremer Pigments, Inc. SAN25 Tyril 100 was obtained from the Dow Chemical. This polymer was chosen as all three dyes are soluble in it to at least 2 to 3 wt. %. SAN25 was first dissolved in dichloroethane at a concentration of 10 wt. % and an appropriate amount of dye was added to the solution, depending on the desired molar concentration (between 0.022 mM to 0.044 mM). This solution was then spun coat on to clean glass slides, and the films annealed at 100 °C for 10 min. This yielded 500 nm to 700 nm thick films, which were then used for all the experiments.

The FL was attenuated in localized spots using single pulses from a 488 nm wavelength, 200 mW CW laser (Omicron LuxX+) externally controlled with a National Instruments DAQ card. The laser was focused on to the sample using a Leica HCX Plan Fluotar, 100×, 0.90 NA objective, which provided a FWHM of around 380 nm in the radial direction and 1.2 μm in the axial direction. Several FLA spots were made on the same sample by translating it along the focal plane of the objective. Spots were made on different dye samples at various pulse lengths and laser power. Samples having different dyes had the same molar concentration. The pulse lengths used to produce the spots were less than 1 μs. Experiments with a longer pulse length (>200 μs) were also performed with laser powers up to 9 μW to consider the non-thermal effects.

The spots were read in a Leica SPE scanning confocal microscope using a 488 nm wavelength laser and an ACS apochromatic 100×/1.4NA oil-immersed objective. The magnitude of FLA is calculated as the intensity of the background FL minus the FL of the written spot, divided by the background FL. The width of a single spot was obtained by fitting the FL profile of the spot to a Gaussian. The data points plotted in the Figs. 2, 3, and 6 are an average of the results from 5–10 spots, written using the same laser pulse parameters, but in different areas and different samples. The uncertainty in all the FLA data points is approximately ±0.03.

Atomic Force Microscopy (AFM) imaging of the spots was performed on a PO-doped SAN25 thin film at a concentration of 0.033 mM. An Agilent 5500 AFM was used to image the spots written at two different zoom levels. To obtain a quantitative topographical image for the bits, scans were performed with a resolution of 512 pixels and a scanning speed of 0.1 lines/s over an area of 44 × 44 μm and 17 × 17 μm. False color for the images was provided using the open source software Gwyddion 2.43, a scanning probe microscopy data visualization and analysis tool.

The determination of the physical causes for FLA relies on an estimation of the temperature of the film. However, it can be difficult to measure the temperature rise directly over the small laser spot, and so simulations are performed. The temperature rise in the dye/polymer system caused by the laser pulse exposure was estimated using COMSOL Multiphysics simulations. The problem of temperature rise by laser radiation incident on an absorbing material has been well-studied.²⁰ Solutions can be found for specific laser intensity distributions and sample boundary conditions,^21,22 but we chose to use COMSOL for accuracy. In general, the temperature rise in a polymer film due to the absorption of light can be determined by solving the heat equation

where T is the temperature, κ is the thermal diffusivity, and Q is the net energy per unit volume per unit time generated within the solid from the laser.

For the simulations, the film was modeled as 8 μm by 8 μm square film, 700 nm thick, with infinite boundary conditions on the edges of the film. Thermal radiation into air was allowed only from the top surface of the film, to simulate the actual spun coat films.

A Gaussian heat source was placed at the center of the film

where α is the absorption coefficient of the sample, t_exp is the exposure time or pulse duration, w(z) spot size of the beam as a function of distance from the beam focal plane, given by

where z_R and w₀ are the Rayleigh range (380 nm) and diffraction limited spot size of the beam, respectively, and the middle of the sample is defined to be z = 0. In Eq. (2), Q_peak is the heat energy deposited at the center of the focused spot, which is

where I₀ is the peak intensity of the laser beam, and E_f is the factor of energy deposited as heat. If we define E_laser as the energy of the laser wavelength and E_band as the energy of the FL band, and QY as the FL quantum yield of the dye film, then E_f can be expressed as

Here, (1–QY)·E_laser corresponds to the energy of the photons that do not cause FL, and QY·(E_laser–E_band) is the amount of energy deposited by the photons that cause FL. For each intensity, exposure time, and dye utilized in producing the data in Fig. 2, the temperature rise at the center of the film at the end of the exposure time is simulated, and plotted as the independent variable in Figs. 2(a)–2(c). An example, of the temperature profile in 0.044M PO films at 63 MW/cm² and different exposure times, is shown in the supplementary material.

The heat equation is solved using the Heat Transfer module of the software package. Heat capacity (2.10 J g⁻¹ K⁻¹), density (1.07 g cm⁻³), and thermal conductivity (0.193 W m⁻¹ K⁻¹) used for SAN25 came from the supplier's data sheet (Dow Chemical).

The experiments described in the Writing and Reading section above were then carried out on the thin films of the three dyes. The photoinduced attenuation of the FL was measured by writing a series of spots using the CW laser at various exposures times. The results for FLA versus laser intensity for 300 ns exposure times are shown in Fig. 2(d). This is data for just one particular time, but the dyes exhibit different degrees of FLA. Similar data for all three dyes at a variety of different intensities and exposure times are plotted in Figs. 2(a) and 2(b) (for samples with two different molar concentrations). These times and intensities are converted into temperature rise at the center spot as determined by COMSOL simulations, and the data are plotted versus sample temperature instead of laser intensity. All the dyes now exhibit the same degree of FLA (the data fall on one common curve). The FLA of each dye exhibits a different trend with intensity (Fig. 2(d)), but all follow the same trend with temperature. It is evident from these plots that there is a sharp temperature threshold that does not depend upon the nature of the dye. All of the different dye/polymer films start to exhibit FLA at around 500–700 K (Fig. 2(c)). The similarity between the threshold temperature for the different dyes is discussed below.

Since temperatures used to analyze the data in Fig. 2 were simulated and not measured directly, we want to independently confirm that the FLA process involved here is at least distinct from the usual photochemical bleaching mechanism.^16,24,25 This latter process requires an exposure time much greater than the 700 ns used here because the chemical reactions usually occur from the triplet state, and the intersystem crossing time from the first excited singlet to the triplet is typically on the order of microseconds.^16,26 Previous studies on photochemical bleaching mechanisms were performed with lower intensity and longer exposure times, and so the specific mechanism responsible for the FLA should be different here. We can verify this by using exposure times much longer than microseconds, where the photochemical bleaching should still be observed with no temperature rise.

The FLA observed in PO and R6G films for exposure times from a few hundred ns to 4 ms is shown in Fig. 3. For exposure times of 100–300 μs and 4 ms, significant FLA is achieved at a much lower temperature rises compared to the sub-μs times. At 4 ms, there is approximately 30% FLA with virtually no theoretical temperature rise. The PO and R6G curves at 200 μs (filled and hollow squares) exhibit different levels of FLA. This is because the photochemical bleaching depends on the chemical structure of the dye, not just on the temperature generated. This is also true for the two dyes 4 ms: R6G exhibits around 0.29 FLA on average, and PO around 0.37. These data show that for exposures times less than about 1 μs, chemical processes are suppressed, and the thermal effects dominate. The specific thermal effects present are described below. Note that the presence of this effect should be largely independent of the laser intensity, as long as the exposure times is less than the intersystem crossing time, the triplet population will remain small.

As the temperatures simulated by COMSOL are well above the glass transition temperature of the SAN25 matrix (380 K),²³ the surface of the FLA bits was also examined by AFM. Figs. 4(a) and 4(b) show the AFM and FL images of spots written at 90 MW/cm² and different exposure times in PO-containing films. For less than 300 ns (Fig. 4(b)), written spots exhibited a slightly raised bump, and an increased FL intensity. This suggests that the temperature rise is large enough to cause the polymer to deform, but the short time for heating limits the degree of melting or FLA.

For exposure times longer than about 300 ns, the written spots are characterized by a pit surrounded by a raised ring, suggesting an apparent flow of the polymer away from the center of focused beam due to a higher temperature over a larger area. A small degree of FLA (less than 10%) is observed at the same times where pit formation begins. Comparison between the pit depth and FLA is shown in Fig. 5. The depth of the pits for times less than 1 μs is a few hundred nanometers, but as the film is around 700 nm thick, the pits only partially penetrate the film. For 1 μs and 2 μs times, the FLA is nearly complete (over 80%) and the pits almost completely penetrate the film (given uncertainties in the ability of the AFM to measure deep pits and the dark counts of the FL measurement setup).

The COMSOL calculations of the temperature rise are only theoretical, but the observation of physical deformation of the polymer suggests that experimentally the polymer is at least reaching its glass transition temperature (380 K). The strong correlation between the magnitude of the FLA and the pit depth (in Fig. 5), as well as the raised ring, suggest that mass flow plays a significant role. It is also possible that the FLA is due in part diffusion of the dye and a direct thermal decomposition of the dye. Since the FLA is correlated with pit formation, thermogravimetric analysis (TGA), which measures the weight loss with temperature, was done to determine the degradation temperature of the dye/polymer systems (supplementary material). For the neat and dye-doped polymers, this is around 700 K. Since we are achieving larger temperatures than this, ablation as well as degradation likely play a role in the FLA. The similarity of the decomposition and glass transition temperatures between the different dye samples is likely the reason for the uniformity in the threshold temperature for the different dyes.

We can look more closely at the effects of thermally induced mass flow and degradation by examining the spot sizes. The size of the FLA spot is calculated by fitting the measured FL profile to a Gaussian. The FWHM of the spot versus exposure time for PO at 0.033 mM and 90 mW/cm² is plotted in Fig. 6. This experimental FL FWHM is compared to the width of the temperature distribution at the threshold value of 700 K, as well as the FWHM of the pits in the AFM. At exposure times less than about 500 ns, the size of the FL profile corresponds more closely to the temperature profile than the AFM profile. For longer times, the widths of all three profiles are similar. These results suggest that the FLA mechanism is as follows. For times less than about 200 ns, there is enough heating to cause some slight deformation of polymer (bumps in Fig. 4(b)), but not enough for outflow. For slightly longer times but below 500 ns, a direct thermal degradation of the dye reduces the FL. The high temperature also melts the polymer slightly, but significant diffusion of the dye is still limited. For times more than 1 μs, a multitude of effects are occurring (diffusion, degradation, and possibly ablation). Understanding the precise degradation products of the polymer or the dye would require an additional investigation.

To develop an ideal dye/polymer system for ODS, we want a dye that can be written at high intensity but will not photochemically bleach at a low intensity and long time scales while reading or storing the disc. It was seen in the FLA experiments that all three dyes showed same response at sub-μs exposure, meaning that all dyes can be written equally well if a high enough temperature can be reached. However, as seen in Fig. 3, the behavior of different dyes deviated from each other in the low power, long exposure time regime where temperature is not a factor. This behavior is more dependent on the properties of the dye and is due to photochemical degradation. The results here indicate that we can find a dye which is very stable under low power corresponding to photochemical bleaching, but that is still writable at high powers. This allows a significant degree of freedom in choosing a dye that is sufficient for both reading and writing of data.

In conclusion, we have shown that a photothermal process can be used to write data in a simple dye/polymer system, leading to a threshold response similar to commercial ODS media. The threshold response limits crosstalk and allows writing below the diffraction limit. The temperature required for FLA is independent of the dye. We also showed that the FLA mechanism shifts away from photothermal to photochemical at long exposure times as expected. This indicates that the ambient photostability is independent of the ability to write data with nanosecond pulses.

Temperatures during the writing process as predicted by the COMSOL Multiphysics simulations and the topographical AFM images indicate a significant role for mass flow as well as thermal degradation of the dye in the observed FLA. In multilayer films, we expect differences due to the fact that the active layer is surrounded by the polymer buffer layer. This will alter the heat flow and thus change the writing conditions. In addition, one might expect that both active and buffer polymers could flow so that, instead of pits and bumps, physical distortions would be observed. Direct thermal degradation and dye diffusion are also possible. Studies of the mechanisms in multilayer films are underway.

See supplementary material for details on extinction and fluorescence of the dyes used, thermogravimetric analysis of the films used, and synthesis of TK01093.

The authors are grateful for financial support from the National Science Foundation Center for Layered Polymer Systems (CLiPS) under Grant No. DMR-0423914.