Flexible, planar, polymeric thin film structures have received much attention as possible components of novel optical and photonic devices due to their tailored functionality, ease of processing, and amenability to large-area, low-cost fabrication.[1, 2] In particular, multilayer polymers, are being developed for filters, sensors, switches and optical limiters, data storage media, and, as highlighted in this review, lasers.
Typically, lasers combine three distinct system features, (i) an emissive gain media, which has an electronic structure suitable for amplified spontaneous emission (ASE) (such as a fluorescent dye), (ii) an optical resonator to enable the feedback necessary for stimulated emission and to control the spatial and spectral coherence of the beam (such as the spaced mirrors of a Fabry-Perot cavity, with at least one also serving as an output facet), and (iii) a pump source (such as another laser, flashlamps, light emitting diodes, or, if practicable, an electrical current) that excites enough electrons in the gain media into higher energy states so that stimulated emission can ensue. This paper is focused on one such type of micro-resonator laser design, the multilayer distributed feedback (DFB) polymer laser. In DFB lasers (first called “mirrorless” lasers), the first two aspects, the feedback mechanism and gain media, are integrated and distributed throughout the structure. In DFB lasers the circulation of ASE, or feedback leading to stimulated emission, within the structure arises from interference due to multiple reflections (diffraction). Schematics comparing two planar designs, a Fabry-Perot laser with selectively reflective end mirrors and the multilayer DFB laser are shown in Figure 1(a,b), respectively. Note that the mirrors in a Fabry-Perot laser may themselves be made from multilayers, as in distributed Bragg reflectors, leading to lasers called Distributed Bragg Reflector (DBR) lasers. In a DFB laser, the mirrors and the gain media share functionality, making possible a compact “micro-resonator.” Beyond their compactness and simplifications in fabrication, DFB lasers require essentially no postfabrication alignment and can more naturally operate in a single longitudinal mode, thereby producing narrow linewidths and spectral/temporal control. As we discuss in detail below, the distributed feedback design also enables tuning of the wavelength to the extent that the multilayer spacing, effective refractive index, and phase relationships can be adjusted in situ.
This review focuses on planar polymer multilayer DFB lasers. The planar geometry allows for large area devices that can be easily optically pumped. Longitudinal, or near longitudinal pumping allows for in situ control of the spatial profile and modes. Large cross-section lasers, for example, might find new applications in displays, reconfigurable laser arrays, laser surface appliqués, and sensors. The high degree of postprocessing flexibility also allows for simple concepts in tuning and beam control. We describe simple postprocessing techniques to lower thresholds and control the laser wavelength. Here, we will emphasize recent results involving a particularly flexible, low-cost, continuous roll-to-roll co-extrusion process. Before describing these results, we briefly comment on a number of alternative approaches to organic DFB lasers.
OTHER TYPES OF ORGANIC DFB LASER STRUCTURES
Corrugated Surface Diffraction Grating Lasers
The most widely studied polymer distributed feedback designs are based on the use of corrugated surface diffraction gratings. Several excellent and very recent review articles and a monograph have appeared on these surface corrugated DFBs and the semiconductor polymers that are typically used in their construction.[11-14] A brief comment on the surface corrugation patterning techniques serves here, not as a complete review, but as a point of comparison with multilayered DFB lasers and their construction by the co-extrusion method discussed in more detail below.
Surface corrugated DFB lasers are constructed as patterned waveguide structures. As such, their construction requires the fabrication of patterns in at least one side of the guiding surface that are on the order of one-quarter the wavelength of the laser output (see Fig. 2). Microfabrication techniques required for patterning submicron surface features date back to the first uses of holographic photolithography, but were limited to the use of photoresists and required a multistep process and a rigid flat surface. Over the past two decades, a wide variety of new techniques have been developed that have improved versatility and simplicity, such as e-beam lithography, embossing,[15, 17] and imprinting, or replica molding. So-called soft- and nano-lithography[20, 21] techniques can be used with elastomeric materials to stamp, mold, and otherwise micro-imprint the desired surface grating.[22, 23]
In soft lithography, a surface pattern from a previously patterned hard master is transferred first to an elastomeric soft stamp, (typically polymerized in situ on the master), which can then be peeled directly off of the master. The elastomeric polymer film, such as poly(dimethyl siloxane) (PDMS), is then amenable to repeated use as a patterning stamp for transfer to a conjugated polymer film either by first coating the stamp with a suitable solvent, or by pressing the stamp directly into a polymer solution followed by solvent evaporation, or by heating the polymer film to be patterned well above its glass transition temperature, thereby enabling it to flow into the elastomeric stamp. Nanoimprint lithography (NIL) is similar except that it typically involves a direct transfer of the pattern from a hard master to a conjugated polymer surface typically using one or a combination of heat, pressure, solvents or UV assisted photo-curing. These techniques have proven to offer high throughput, once a suitable mold and processing parameters have been established. Room-temperature, pressure techniques have been adapted for use particularly with low molecular weight oligmers. Solvent-assisted techniques are especially well-suited to organic semiconductor conjugated polymers which degrade under the higher temperatures required for melt processing.
Nanoimprinting techniques are also well-suited for small area patterning directly onto a suitable optical pump surface such as a laser diode, LED, or, it is eventually anticipated, electrical contacts. Still, these techniques are useful primarily for waveguide laser geometries, not for deep patterning beyond the surface layers. They require a suitably patterned master that is usually not amenable to adjustment of the grating period during the embossing process. Also, these techniques are typically not solvent free, although there appears to be no inherent reason why these processes cannot be applied to a neat polymer film from melt-processing, for example. Further, to the extent that corrugation techniques rely on elastomeric materials that are amenable to the soft lithography printing processes, they have been troubled by distortions. Surface patterning techniques, while promising on many fronts, may not be as amenable to very large-area, mass production as roll-to-roll co-extruded multilayers.
Perhaps the most significant distinction between the multilayer polymer DFBs to be reviewed here and surface patterned DFB lasers is found in the output coupling of these two laser types. Multilayer DFB lasers are true surface emitting lasers whereby the beam characteristics, discussed further below, are determined by the short cavity lengths and mode structure. The first surface corrugated DFB laser structures, on the other hand, produced waveguide edge emission. High optical quality cleaved edges needed for edge emission are difficult to achieve in amorphous polymers. Also, the output beams are highly asymmetric, as expected from the waveguide structure. One way around the issues with edge emission has been to use second order grating effects to couple the light to surface emission from the grating. While pure second order gratings tend to require higher thresholds for emission, typically by a factor of two, recent efforts have overcome this problem by using a combination of a long first order grating pattern to produce the distributed feedback, but with a short segment of no more than a few tens of periods of second order grating at the center to introduce a phase matched output region for surface emission, as illustrated in Figure 2. Thresholds as low as 57 W/cm2 have been reported using this mixed order grating technique.
For the remainder of this review, we focus primarily on a multilayer design that creates a two-dimensional surface for lasing, rather than edge emission or higher order surface emission common to surface patterned grating systems. The melt-processed multilayers described and studied here are readily scalable for mass production of multilayer interference mirrors and DFB lasers, even over very large surface areas. These materials can form two-dimensional surface-emitting array lasers for parallel processing systems. For co-extruded multilayer DFB laser systems, there are exciting prospects for commercial applications due to their simplicity, low-cost, ease of processing, and flexibility as they continue improving toward the goal of matching or exceeding the lasing performance, low thresholds, and narrow linewidths of corrugated grating based DFB systems and other microresonator lasers.
Holographic Interference in Dichromate Gelatins
While holographic interference lithography techniques have been commonly employed to produce corrugated gratings for DFB lasers, it is worth mentioning that there are also specialized interference techniques that produce layered systems. The holographic interference layering process has, to date, been demonstrated only in high-resolution dichromate gelatin (DCG) emulsions, but has been successfully used to create simple DFB lasers and defect DFB lasers by superposition of two interference patterns with overlapping band edges. In DCG systems, the layered structure results from the interference of a shorter wavelength writing laser, dye diffusion into the gelatin through a swelling process followed by dehydration and baking and sealing. Resulting refractive indices created by the interference were on the order of 1.41 to 1.5, requiring a large number of layers for a high quality reflection band. Interestingly, in Ref. , the final emulsion layer spacing is not uniform throughout, but is graded in thickness, i.e., the layer spacings are narrower on one side of the film than the other. Graded multilayers have been shown to result in a widening of the bandgap,[31, 35] and have recently been used to enable multi-wavelength DFB laser arrays.
In addition to the polymer forced-assembly technique described below, several mechanisms for self-assembly are being explored, such as the use of block co-polymers or co-polymer/homopolymer blends.[37, 38] Self-assembly of block co-polymers has been used for the multilayer step for distributed Bragg (DBR) reflectors, but requires specialized synthesis steps to obtain the desired layer thicknesses.[39, 40] Use of self-assembly for DFB lasers is more restrictive also because the process does not lend itself to large refractive index differences, large areas, and simultaneous optimization of the gain media in alternating layers of the self-assembled structure.
Chiral Liquid Crystals
Though not strictly all-polymer (but see Refs. [41-43]), a great deal of work has been done in the area of chiral liquid crystal DFB lasers,[44-47] wherein the multiple interference feedback structure is achieved through the spontaneous assembly of chiral nematic or chiral smectic periodic (helical) structure. Intermediate phases between smectic and nematic, called blue phases, are also found to form periodic structures with reflection bands. For a review of the tuning response of cholesteric liquid crystals, see Ref. .
The rest of this review is organized as follows. It is useful to relate the phenomena we describe below through a simple theoretical framework, so we first provide a brief tutorial on the theory of multilayer DBR and DFB laser microresonators along with a review of the gain media used in these systems. We then describe forced assembly techniques used in fabricating multilayer polymer DFB lasers using melt-processed co-extrusion. We serially review work that exploits the versatility of these systems, focusing particularly on novel techniques for postprocess tuning of multilayer polymer DFB lasers. Finally, we speculate on the prospects for this type of laser for diverse applications.
MULTILAYER CAVITY DFB DESIGN/THEORY
One-Dimensional Photonic Bandgap Effects
In this section, we review some of the basic optical properties of the DFB structure in both band-edge and defect-mode lasing configurations. As first proposed and demonstrated at Bell Labs in 1971 by Kogelnik and Shank, the DFB laser requires a periodic variation of either the refractive index or the gain profile or both. This periodic index alternation (which need not be strictly bimodal and may include significant regions of gradient refractive index between the layers, mixed grating orders, or even be sinusoidal) produces a reflection band over a range of frequencies. A binary multilayer DFB resonator consists of alternating layers of two materials of contrasting refractive index , only one of which contains the gain media, in approximately quarter-wavelength optical thickness increments for j = 1,2 and odd integer values of m [see Fig. 1(b,c)]. The first-order free-space wavelength center ( ) and spectral width of the reflection band of such a system of two materials with real indices and are thus given in the simple binary DFB system by:
To model the resonator properties associated with a multilayer interference, one commonly represents each layer and interface by a characteristic transfer matrix describing the transmission/absorption/reflection through each layer/surface. The structure's overall transmissive, reflective and absorptive properties (at normal incidence in the absence of birefringence or other polarization anisotropy) can be determined by the product of these transfer matrices, modified, if appropriate, to account for loss of coherence. (If birefringence is significant, the resonator structure can be modeled with matrices; see, for example, Ref. . For simplicity, however, we discuss the case of polarization preserving transport only.) Briefly, in the propagation basis, the linear relation between the amplitude of the fields in successive layers can be found using the matrix product to represent successive layers j of thickness dj and the interfaces between them. Thus, the phase evolution associated with propagation through the layer is given by the matrix:
and the interface between the jth and (j + 1)th layer is represented by:
where we have used the Fresnel (complex) reflection and transmission coefficients (both given at assumed normal incidence).
In each matrix, the index of refraction is treated as a complex function of the wavelength, and can parameterize either absorption loss or gain in the jth layer. (For our present purposes, we ignore the gain dependence on the pump parameters, hysteresis, and other factors related to pulsed pumping.) The total system transfer matrix for N layers is thus:
and, for example, the overall transmittance of the laser resonator is simply expressed by .
The foregoing method models the theoretical transmission spectra for representative multilayer systems as shown by example in Figure 3. Consistent with photonic crystal (PhC) terminology, the band gap is the region of low transmissivity (high spectral reflectivity). As predicted by eqs (1) and (2) and seen in Figure 3(a), low refractive index contrast leads to a narrower band gap and the need for a larger number of layers to produce a higher contrast (sharper-edged) reflection band.
Extension of this transfer matrix technique to non-normal incidence is straightforward, but requires separate treatment for TE and TM incident light. Details of the calculation are left to the references, but it is important to note that at non-normal incidence, regardless of polarization, the bandgap shifts towards shorter wavelengths. This fact becomes particularly relevant when optically pumping DFB lasers if there is only a small separation between the absorption and emission peaks of the gain media (small Stokes shift). One can use this angle tuning of the bandgap to increase the absorption of the pump light by shifting the reflection band away from the pump wavelength. Further studies of the effect of angle of incidence of the pump beam on a similar system can be found in Ref. , which also explores differences between low and high frequency band edge lasing.
An interesting effect occurs when the periodic multilayer system is folded to create a defect (doubled) layer in the center of the stack [see Fig. 1(c)]. Figure 3(b) shows that the bandgap splits and a narrow transmission region appears near the center of the bandgap. This break from perfect periodicity in the multilayer is an example of what is called a “phase-slip defect” which results in one or more transmission defects in the reflection band.[55-58] In a simply folded binary DFB system, the phase slip defect naturally appears as either a thicker half-wavelength region of low refractive or high refractive index in the center of the stack. Note that repeated folding or breaks in the periodicity of the structure lead to additional defect modes, which may be useful for fabricating multiple wavelength lasers if mode competition issues can be addressed. Ref.  also discusses the theory of cascaded phase-slips in a DFB resonator. For a single fold, differences in the resulting DFB laser behavior for these two different fold directions result from whether gain occurs in the low index or high index material. The performance of the resulting “defect” DFB laser is also strongly dependent on the location and width of the structural defect in the stack. As will be discussed below, some structure defects (for example, due to variations in the layer thicknesses) may not be completely avoidable, but their presence may also be advantageous for lowering the threshold and enabling tuning of the DFB laser.
Many excellent monographs are available for a more detailed discussion of the band structures generally of photonic crystals, with and without defect states. In addition to the citations above, see, for example, Refs. [60, 61] to .
Group Velocity Delay and Modeling with Gain
To connect the properties considered thus far with the physics of a DFB laser, it is instructive to consider how the multilayer microresonator structure effectively slows the propagation of light, increasing the light interaction with the medium and, equivalently, increasing the electric field energy locally in the multilayer. To understand how the structure will respond to the introduction of a gain media and pump source, it is useful to first calculate the density of states or, equivalently in one-dimension, the inverse group velocity . The group velocity can be obtained directly from the inverse slope of the phase retardation of the system. See Figure 3(c) and, for example, Ref.  for a derivation of the group velocity in a bilayer system with full Bloch eigenfunctions. In an infinite DFB laser with uniform layers, the group velocity vanishes as a power law at perfect band edges, implying increased light/matter interaction and consequently the potential for large gain. The rate of spontaneous emission at a particular frequency, using Fermi's Golden Rule, is proportional to the density of states at that frequency.
To calculate the properties of the full DFB laser with gain, a variety of approaches can be used, such as coupled-wave theory and plane-wave eigenmode expansions. One particularly simple approach, taking advantage of the matrix formalism developed so far, is to assume a negative imaginary part to the refractive index representing gain. A more careful approach will use the finite time propagation of the pump pulse and the time dependence of the lasing field. In Dowling et al., the effects of gain were determined directly by solving for pulse propagation according to the Maxwell wave equation,
where n(z) is the refractive index and g(z) the gain function throughout the stack. Practically, the solution to this equation for a finite structure is accomplished through a finite difference, time domain (FDTD) numerical technique. Details of the FDTD calculation can be found in Ref. . In this method, light emitting sources are simulated inside the stack to mimic the distribution of light emitting gain media (dye) inside the structure. The resulting light field is then allowed to propagate in accordance with Maxwell's equations in a step-wise fashion. The limitations of the method are derived from the need for small time steps and a fine spatial grid, and one must take particular care at the discontinuous dielectric interfaces, e.g., layer boundaries. This method can be computationally intensive, but there are efficient, readily available, numerical packages that facilitate these calculations.[72, 73]
Regardless of the method used, the periodic index variations of these multilayers leads to a pulse-limited, quasi-standing wave inside the film, with the buildup of field intensity inside the film enhanced several times over the maximum input field intensity. Effectively, the structure is designed to use interference to pile up energy in the antinodes, where it can be absorbed by the gain medium there. The location of these antinodes is a function of the wavelength, appearing in the high refractive index material on the long wavelength edge of the bandgap and in the lower refractive index material on the shorter wavelength bandgap edge. Thus, whether the lasing threshold is preferred at the high-energy or the low-energy edge of the reflection band depends upon whether the gain medium is in the lower refractive index or higher refractive index constituent, respectively, as illustrated in Figure 4.
Because any gain medium is also an absorber of pump energy, for the lowest lasing threshold it is preferable to include gain media in only one of layered materials, with the choice depending upon where the peak amplified spontaneous emission (i.e., the standing wave antinodes) most closely overlaps a bandgap edge or defect mode. The center of the broad fluorescence spectrum of the gain functionalized polymer is matched to the bandgap edge/defect of the multilayer. Because spontaneous emission is inhibited in the bandgap, amplified spontaneous emission appears preferentially at the bandgap edges or at defects in the bandgap. Similarly, at low pump intensities the fluorescence emission is suppressed in the reflection band, but becomes enhanced near the band edge or at band defects. Nonlinearity of stimulated emission implies that, as the pump intensity increases, the spectral width of the fluorescence at the band edge or band defect narrows continuously through the lasing threshold.
Although theoretical expectations are that perfectly regular DFB lasers lase at the band edges, experimentally, lasing typically occurs at wavelengths where random layer thickness variations or phase slip defects have enhanced the density of states (i.e., spectral neighborhoods of lower group velocity). Furthermore, the presence of disorder tends to decrease gain at the band edge, increasing the corresponding lasing threshold. In a disordered multilayer, is expected to vanish logarithmically (e.g., much slower than power law) with the disorder parameter—which scales as the inverse of the number of layers, 1/N, for a finite system. In a similar finding, a randomly amplified layered system has been shown to lase at localized modes due to slowing of light at those modes. Lasing at well-defined defect modes, however, has been shown to lead to lower threshold lasing. Even if the defect layer is much thicker than a quarter-wavelength, one expects enhanced lasing at a defect state, which corresponds to a slower group velocity.
Figure 5 expands on Figure 4 by showing the results from adding broadband (flat) gain to either constituent of the multilayer stack and folding onto that or the other constituent, respectively. As before, the highest peak in the gain profile shows the spectral location expected to exhibit the lowest threshold for lasing. Note, however, that the presence of other peaks shows that there is competition between modes in the structure and, well above threshold, lasing may occur at different wavelengths or even simultaneously at multiple wavelengths.
Although we leave the details of the calculation to Ref. , it is straightforward to use the transfer matrix technique to map out the electric field intensity as a function of location within the multilayer structure. Doing so, one finds that the defect structure yields a peak internal field energy density at the spatial location of the structure defect and at the wavelength of the spectral defect. This field energy density at the defect location (center fold) exceeds that found at the band edges of a simple stacked structure with the same number of layers, where the field energy density is greatest spectrally along the band edge, and is less localized spatially within the film. As the number of bilayers is increased, the peak field energy density is both further enhanced and more localized at defect states as compared with multilayers without defects. Figure 6 shows the results of numerical transfer matrix calculations of the electric field energy densities in hypothetical folded 32 layer films for an incident field from the left. In case (a), the center region is the lower refractive index material (here, n = 1.49) and in case (b) the center fold is on the higher index material (n = 1.585). The regions of high energy density correspond to the antinodes referred to above and the energy is highest in the center of the fold, by a factor of two or more as compared with when the low refractive index material is in the center. Thus, it is expected (as in Fig. 5) that a defect binary DFB laser is optimized by having the gain medium in the lower refractive index material, with a doubled layer of low index material at the center of the stack.
Having identified the basic structure and the optimal placement of the gain media for a DFB laser, we now briefly consider the nature of the gain media particular to fabricating polymer multilayer DFB lasers. The earliest organic laser gain media, first developed shortly after the laser's invention, were fluorescent dyes,[79, 80] first used in liquid solutions, mostly with toxic solvents, and with the commensurate hazardous waste problem. It was not long, however, that these dyes were first adapted for use in a solid polymer matrix,[81-83] commonly poly-(methylmethacrylate) (PMMA). Not surprisingly, most multilayer DFB lasers have also taken advantage of the wide range of fluorescent dye molecules that are amenable to being dissolved into or chemically attached to the polymer host or dispersed with nanoparticle composites. These dyes are typically π-conjugated molecules with high quantum fluorescent yields, with some of the most common representative examples being the xanthenes (rhodamine [see Fig. 7] and fluorescein dyes), pyrromethenes, coumarins, and, more recently, organic photovoltaic (OPV) chromophores and semiconducting polymers, with optimal ranges for lasing from the blue to the near infrared.[8, 84-89]
To determine the appropriate gain media concentration requires consideration of the interaction length in the multilayer. The actual thickness of gain material may be only on the order of ∼10 microns (based on ∼100 dye-doped layers at ∼100 nm each), but the interaction length with the dye is much longer due to multilayer reflections. If the interaction length is sufficiently short that large concentrations are needed, then intermolecular interactions become very important and often lead to quenching, which is detrimental to the lasing efficiency and stability. For low intensities, as expected, the light intensity increases nearly exponentially with the interaction length in the material.
where is the stimulated emission cross section and is the population inversion density in the material, which together constitute the gain, .
In their recent review of solid-state organic lasers, Chénais and Forget provide a concise listing of the key features needed for the gain media in a polymer laser:
Stability against moisture and oxygen, or encapsulation in an impervious barrier to the same.
Photostability at the pump photon energies.
Large quantum fluorescent yield and low-quenching in the solid matrix.
Low re-absorption or scattering losses at the lasing wavelengths.
Large stimulated emission cross-section to enable low thresholds.
Low triplet-triplet absorption, low triplet state lifetimes, and a low intersystem crossing rate.
One of the important characteristics of many laser dyes is the breadth of their emission spectrum, which is a prerequisite to tunability, but can also raise the lasing threshold. Below, we show experimentally and theoretically how multilayer co-extruded polymer DFB lasers take advantage of a dye's broad gain envelope through several different tuning modalities. The wide spectral range of the gain envelope also is responsible for the ability of these dyes to be used for generating short duration pulses.
The photoluminescent efficiency of light emission is described in terms of the quantum yield, the ratio of the number of photons emitted to the number of photons absorbed. Unfortunately, as the concentration of dye or active conjugated molecules increases, many otherwise promising gain materials tend to form aggregates, dimers, or excimers. Dyes that are strongly emissive in liquid solutions are easily quenched by these interactions in a solid matrix. To prevent this quenching, it may be preferable to attach the dyes as carefully spaced side groups to avoid aggregation. (This problem has also been considered extensively in the literature on organic light emitting diodes (OLEDs) with some success through the use of dendrimers to space out the chromophores from each other.)
Unfortunately, in solid matrices, dyes used to date have tended to suffer from low efficiency and fast photodegradation as compared with their liquid-dissolved forms. The process of photodegradation in systems of organic molecules has been studied for over half a century. Optical devices that use organic dyes have a limited lifetime, as short as seconds to weeks, before the dye or entire device must be replaced.[97-99] Although the lifetime of some organic dye-doped systems has been extended in certain cases, current understanding of photodegradation is still far from complete. The reasons for the lack of a complete theory are twofold: (1) the large assortment of dye configurations with differing properties, and (2) the varying routes of photodegradation in each species of dye, such as chemical oxidation, formation of dimers, and basic thermal degradation. Examples of photodegradation mechanisms include phototautomerization, photoisomerization, photodenaturation, photoejection,[103-106] triplet-radical reactions, photodimerization, photodissociation, and twisted intermolecular charge transfer (TICT). Some of these processes such as photoisomerization are known to reversibly recover, while other processes like photoejection only show signs of self-healing when, under the right circumstances, both charge trapping and recombination occur. Within the current literature, however, most processes of photodegradation are irreversible with many molecular products of photodegradation still unknown.
Work on modification of laser dyes to make them more suitable for solid state applications is ongoing. For example, the dye commonly known as DCM (4-(dicyanomethylene)−2-methyl-6-(4-dimethylaminostyryl)−4H-pyran) has been modified by mixing it in a guest-host system with Alq3 (tris(8-hydroxyquinolinato)) instead of a completely passive matrix like PMMA. The combination not only helps to separate the DCM dye molecules, limiting concentration quenching, but it does so by simultaneously increasing the efficiency of the fluorescence process because the Alq3 molecule absorbs the pump light, and efficiently transfers that energy to the lower energy gap DCM molecules. This serves to absorb pump light more efficiently, and yields a larger Stokes shift between the absorption and emission wavelengths, reducing self-absorption. Note that in single species systems, a small Stokes shift is desirable in order to reduce the amount of pump energy converted into destructive thermal energy in the system, but a large Stokes is desirable to limit re-absorption. The Alq3-DCM system thereby suggests a possible compromise between these two competing interests, yet issues relating to various routes to quenching in these systems persist.
Another approach is to use organic semiconductor (OSC) polymers which are undoped, conjugated fluorescent polymers,[113, 114] such as poly(phenylene vinylene)s[115-118] (PPV) and polyfluorenes,[119-121] Other examples of materials with these properties included conjugated dendrimers and spiro compounds, which are oligomers coupled by spiro linkages. Polythiophenes have also been shown to be low threshold emitters in microcavity lasers. To further lower thresholds and to facilitate short wavelength pumping, several groups have also used combinations of OSCs and dyes.[125-130]
OSC gain media offer two primary advantages over their dye-doped polymer counterparts: (1) because they do not suffer as significantly from concentration dependent quenching, except by interactions between polymer chains, they can be used at much higher concentrations and so may even be used without dilution, enabling large photoluminescence quantum yields (PLQY) and short cavity lengths. (2) OSCs are capable of charge transport, and so offer the possibility of electrical pumping. The conducting pathway, however, is not available in diluted OSC solutions or if the conjugation chain is broken through oxidation or other degradation mechanisms. At this writing, the more likely scenario for integration of low threshold DFB lasers into devices is through mating the DFB laser to a low energy laser diode or LED pump. Significant progress has already been made in this area using corrugated surface DFB structures.[23, 30]
Unfortunately, OSC lasers have only been fabricated by solvent or vacuum deposition techniques so that planar multilayer DFB structures have not yet been reported. OSCs are not amenable to melt-processing techniques due to degradation effects at elevated temperatures. Also, the requirements for matching melt viscosities of polymer pairs in co-extrusion further precludes the pairing of OSCs with other suitable optical polymers.
As can be inferred from this brief review, much work is being done to improve gain media for all-polymer lasers. The range of laser dyes used in multilayer polymer DFB lasers to date is rather narrow. Particular examples are discussed below in context of reviewing laboratory systems to date. For further discussion of the advances in organic/polymeric laser materials generally, there have been many helpful reviews to which the reader may refer.[11-13, 84, 93, 131-134]
There are a wide variety of metrics used to characterize lasers and laser materials and a complete overview of parameters is not only beyond our scope, but is also made problematic by the varied approaches to pumping these lasers. In the discussion of experimental implementation, we routinely quote the lasing threshold in terms of either pump irradiance or fluence (average pump energy divided by the pump beam area) at the onset of lasing. Also, we quote the lasing quantum slope efficiencies in terms of the steepest slope of the so-called “J-curve” of output power vs. pump power, as shown in Figure 8(a). Quantum slope efficiency differs from the slope of the steep part of the curve shown only in that it is expressed in terms of the number of photons output per number of photons in the pump, i.e., the slope shown multiplied by the ratio of the pump wavelength to the lasing wavelength. The threshold, however, must be inferred by knowing the spot size of the pump laser at the pump power where the curve changes slope (here, at <10 µW pump power).
In our experimental work, further described below, we find that at threshold the output undergoes profound spectral narrowing, as expected on general grounds in lasing. Additionally, the coherence of the output light is responsible for several features in the observed output beam. Figure 8(b) shows the output beam from a co-extruded polymer DFB laser (vide infra). Notable features include the innermost cone of output light and the rings from etaloning inside the laser cavity. The emission cone is consequent to both diffraction from a small laser output facet, or, alternatively, the presence of (transverse) multimode lasing. Although the pump may be Gaussian and the lasing is naively expected to inherit the coherence properties (both spatial and temporal) of the pump, the laser is typically pumped at an angle to the laser cavity and the inhomogeneities in the laser cavity itself reduce the coherence of the laser output. Even with the angle pumping and the inhomogeneities, the azimuthally symmetric emission pattern observe from these surface emitters [see Fig. 8(b)] simplifies the beam shaping requirements of the optical chain beyond the laser and thus is an advantage these broad area lasing structures have over other types of DFBs. The azimuthal isotropy and the laser geometry also imply that the peak power from these devices is likely to be higher than surface corrugated DFBs.
The short cavity lengths of multilayer DFB lasers ensure that, at single wavelength operation, there is but one longitudinal mode. Transverse modes, however, are more difficult to characterize. Guided wave lasers, such as the surface corrugated DFB systems, offer advantages here in that the waveguide controls the modal volume, helping also to lower the laser thresholds. To further lower thresholds in multilayer DFB lasers, it may be necessary to control the transverse mode structure. One of the ways by which this may be accomplished is, not surprisingly, by surface patterning. For example, one might pattern a surface microlens on the outer layers to create a confocal resonator, or even a concentric circular ring structure, as further described in Ref.  to provide improved transverse mode feedback.
Unfortunately, neither the threshold nor the lasing efficiency can be considered to be completely independent of the pump duration, repetition rate, and other parameters, and comparisons across research reports are not always reliable indicators of the relative performance between systems. Also, the presence of a threshold for increased emission efficiency is not sufficient evidence in the absence of the other conditions for lasing. At threshold, one finds substantial spectral narrowing of the output mode expected for the DFB structure and evidence of coherence in the output beam. A lower bound on the laser coherence length can be inferred by observing the contrast in successive diffraction rings. An upper bound can be inferred from how well the laser pulse mirrors the pump pulse in the time domain. Of course, measuring the degradation of the contrast of an interferogram as one changes the lapse between the beams is the most direct way the measure the coherence length, but for short coherence length sources that can be challenging.
Other metrics to consider are the peak power, damage threshold, and useful life. The latter two metrics are typically a function of the gain medium and the ability of the structure to dissipate thermal energy while efficiently absorbing pump energy. Issues with degradation of dye-doped materials have already been discussed above and represent a significant obstacle to some applications.
Fabrication: Putting it all Together
As illustrated in Figure 3, the number of layers needed for the DFB/mirror system is a function of the refractive index difference between the layered materials as well as the amount of gain in the layers. Large refractive index contrast can be found in hybrid organic/inorganic systems having low numbers of layers and wide band gaps, but these systems are not amenable to easy processing. Large and small refractive index systems (n > 1.7, n < 1.37) typically require more complex manufacturing methods such as vacuum deposition, liquid phase epitaxy, etc. and are beyond the scope of this review. Thus, to achieve the large reflectivities needed for a DFB laser structure with all-polymer multilayer systems typically requires on the order of a hundred layers or more due to the relatively small refractive index differences available.
Once a suitable dye-host or OSC polymer system is found, another issue is the mechanical, chemical, and thermal compatibility with the other polymer making up the DFB and the layering process. Similarity in structure between polymers can lead to large interfacial regions, which can be particularly problematic for maintaining the interface between gain layers and chromophore-free regions. The two techniques used to date for forced assembly are spin-coating, which is a well-established, but more tedious process for producing large numbers of layers, and melt-processed co-extrusion.
An assembly technique that requires constructing the multilayer one layer at a time may be primarily of research interest due to the time and number of steps required to assemble enough layers. Nonetheless, some early work on the construction of distributed Bragg multilayers by spin-coating and other layer-by-layer technique merits discussion. Recent work suggests ways to improve the speed of the process, but which have not yet been applied to a DFB system.
The general process of spin coating of polymers from dilute solution is well known, having been employed in the semiconductor industry for decades in the use of photoresists. Because the DFB laser process usually employs ultrathin films of less than 200 nm, however, the control of the layer thickness and uniformity can seem more of an art, requiring repeated trial and characterization through interferometric or profilometric means, especially when thickness control at the nanometer level is desired. Broadly, the film thickness has been seen experimentally and roughly modeled to scale as
where is the initial solution viscosity, and is the spin rate, with thicker layers requiring slower speeds at the risk of loss of uniformity. This prediction depends, of course, upon many other material parameters such as the mass fraction/polymer concentration in the solution and characteristics of the solvent gas phase above the sample, e.g., the diffusivity of the solvent in the gas phase. Generally, less volatile solvents lead to more uniform films due to slower solvent evaporation. Further, high viscosities, which can arise from high concentration of polymer (more than a few percent), lead to less uniform films. A detailed consideration of the parameters involved in relating the spinning parameters to the resulting film thickness can be found in Refs.  and , which consider initial solvent volatility/evaporation rate during the coating and spinning process, as well as non-Newtonian fluid effects.
The use of spin coating to produce a multilayer polymer system dates back to the 1970s. The quality of spin cast films is typically limited by the number of layers spun which is limited by the tendency for the dissolution or disruption of previous layers as each new layer is added as well as by the time and effort required for each layer. Spin casting of multilayers requires the use of mutually exclusive solvents for the alternating polymer layers. Bailey and Sharp spun 50 alternating layers of polystyrene (PS) and poly(vinylpyrrolidone) (PVP) producing films with 55% reflectance in the visible (a higher order reflection band). Although this reflectance is too low for use as a laser mirror or DFB system, they found that automation of the spin-coat process (primarily rotor speed) enabled them to produce chirped structures useful for customizing the reflection band. The interdiffusion regions at the interfaces were estimated to be around 10 to 20 nm for the spun layers.
Multiple groups have successfully fabricated flexible multilayer DFB lasers by spin coating alternate layers of cellulose acetate (CA, n ∼1.475NaD) (dissolved in alcohol) and poly-N-vinylcarbazole (PVK, n ∼1.683NaD) (dissolved in chlorobenzene).[146, 147] In Ref. , the CA layers were doped with 0.5 wt % R6G, which exhibited band edge lasing at 580 nm in a simple alternating structure of 19 layers. When pumped by 5 ns frequency doubled Nd:YAG pulses at 532 nm and 1 Hz repetition rate, the threshold was estimated to be 17 mJ/cm2. In Ref. , both the CA and PVK polymers were doped with 2 wt % pyrromethene-567. Single mode lasing was achieved at a band defect at 568 nm, created by inserting a single wavelength spacer near the center of the DFB structure. In that case, the total of 68 layers was made in three parts, with a center 400 nm doped CA gain layer, between 31 and 36 dye-doped CA/PVK reflector layers. Compared with a similar DBR system with only a center doped layer, the threshold at 260 nJ/pulse (or 300 µJ/cm2 pulse energy density) was substantially lower due to the doping of the reflector layers. Similarly, the laser was pumped with the frequency doubled output of a Nd:YAG laser at 532 nm at 7 ns pulse duration, with a 10 Hz repetition rate.
An alternative to repetitive spin coating, though still somewhat tedious, is spin coating of large removable single layers that can be then cut and stacked to form the multilayer structure. This is the technique used by Komikado et al. who built an early all-polymer DFB laser in 2006 by spinning CA and PVK at a thickness corresponding to three-quarter and one-quarter wavelength optical thickness, respectively, and stacking them in a 39-layer stack. The CA layers were doped with R6G at 0.5 wt %, which also necessitated that thicker layers be used to ensure sufficient absorption of the pump laser. The resulting laser output was, as expected, at the short wavelength band edge of 590 nm when pumped by a doubled Nd:YAG laser at 10 ns pulse duration. The lasing threshold was ∼50 µJ/pulse.
The versatility of the doped PVK/CA system has also been shown in a hybrid structure comprised of a 10 pairs of TiO2/SiO2 layers forming a DBR mirror on which a 1-µm thick coumarin 540A doped PVK center layer was spun followed by up to 25 pairs of CA/doped-PVK bilayers. This DBR/defect/DFB hybrid inorganic/organic multilayer laser was then pumped by a 4-ns pulsed InGaN-based blue laser diode at 441 nm and a 100 Hz repetition rate. The lasing threshold at 563 nm was found to be 370 µJ/cm2, much lower than the same group's similar effort with pyrromethene 567 doped and pumped at 532 nm by a doubled Nd:YAG laser.
Also of note is the work by Joon et al., in which spun alternating layers of PMMA (99 nm each, n ∼1.49) doped at 0.5% with DCM, and titania (TiO2) nanoparticles (88 nm, n = 1.78 at 500 nm) formed a 61-layer structure with a 1.54 µm dye-doped PMMA gain layer at its center. The assembled multilayer lased at a 582 nm defect in the reflection band (coinciding with the peak emission wavelength of the DCM doped PMMA) when pumped by a frequency doubled Nd:YAG with 5 ns pulses at 50 Hz repetition rate. The lasing threshold was found to be ∼17mJ/cm2 (12 µJ/pulse over a 300 µm pulse diameter).
Multilayer co-extrusion of polymers has been used to make multilayer interference gratings for over 40 years, since first developed at Dow Chemical Company, and it has been used commercially in the production of low-cost, large-area reflective films. A major breakthrough in the creation of low-cost multilayer DFB lasers appeared, however, when new developments in customizing the multilayer co-extrusion enabled increased design flexibility in the multilayer process.[152, 153] In this process, two thermoplastic polymers of differing refractive index are melt-pumped through a series of layer multipliers, splitting and recombining the melt flow with minimal path length differences along the extrusion path. The extrusion temperature used is that for which the rheologies of the two polymers match, enabling them to flow together more uniformly through the system. (A plug flow rheometer or melt flow indexer is used to predict the melt viscosity as a function of temperature.) The operation of this co-extrusion process to enable roll-to-roll processing of multilayer DFB lasers and other devices can be found in Refs.  and .
As can be seen in Figure 9(a), at each layer multiplier the horizontal stack is split vertically into two melt streams and then recombined by forcing half the melt to flow above the other half, doubling the number of layers at each stage. By repeated dividing, spreading, and stacking, the number of layers grows as 2n + 1, where n is the number of doubling dies used. After the desired number of layers has been reached, which can be in the thousands, a removable surface “skin” layer, such as polyethylene (PE), is typically extruded to protect the multilayers, which are then spread in an exit die to form large area films, as seen in Figure 9(b). Note that the skin layers are thick compared with the multilayers and occupy 50 to 90% of the final film volume, enabling easier handing of the films. After the melt has been spread into the desired dimension and corresponding layer thickness, a chilled take-up roll is used to quench the melt and smooth the surface finish. The final multilayer film, after removable of the skin layer, is typically 3 to 12 microns thick overall.
The roll-to-roll technique enables the solvent-free fabrication of multilayers for which the band gap shape and position can be readily modified to the desired design. To date, over a hundred different polymers have been successfully co-extruded by this method, with good optical quality films obtained over a wide range of polymers, including polypropylene(PP), polyethylene (PE), polystyrene (PS), polycarbonate (PC), poly(vinylidene fluoride) (PVDF), PMMA, and related blends, some of which are described further below.
Experimental Characterization of Multilayers
For any multilayer DFB laser fabrication method, it is useful to consider the uniformity of the layers across the area of the film and the layer-by-layer thickness uniformity to the desired periodicity. Unfortunately, standard ellipsometry techniques used to measure thin film thicknesses are not capable of investigating more than a few layers into the multilayer stack. Furthermore, it is a very complex inverse problem to use the features of the transmission or other spectra to infer the actual layer thicknesses, particularly in the presence of absorptive and scattering losses. The most direct approach, then, is to cleave across the multilayer polymer stack, which is most effectively done with a cryogenic microtome, and then scan across the cleaved end using atomic force (AFM) or scanning electron (SEM) microscopies. Figure 10(a) shows the results of one such AFM scan across a 128 layer film described in Ref. . In this case the layers were found to be 95 ± 25 nm thick, but, more significantly, there is good agreement between the transmission spectra predicted by the transfer matrix method using the measured layer thicknesses and the actual transmission spectrum of the film [Fig. 10(b)]. Note that foregoing even just the last layer multiplication step (that is, stopping at a 64 layer film), the layer thickness variation is markedly reduced to about 18%. The co-extrusion system described here offers great versatility for producing a wide variety of multilayer structures. In a commercial system, however, the multilayer co-extrusion process may be accomplished without the use of multipliers, thereby leading to high uniformity, but with some loss of customizability.
Progress with Co-Extruded Multilayer DFB Lasers
The first use of multilayer co-extrusion of microresonator laser cavities was not as a DFB laser, but as distributed Bragg reflectors (DBRs) laminated onto a dye-doped gain layer two or more orders of magnitude thicker than the individual Bragg layers. The best performing of these systems showed a threshold of 35 µJ/cm2 at 50% optical efficiency using a 53 µm pyrromethene gain layer in a PVDF/PMMA blend at 1.9 wt % of dye. This DBR approach has the advantage of versatility in that the center gain layer can be almost any gain media in any compatible host. The disadvantage is, of course, the extra processing required to combine the mirrors and the gain medium. This system is noteworthy also for the insights it provides for our understanding of the effective length of the resonator cavity with DBR mirrors in relation to the thickness of the gain layer and the presence of nonuniformity in the mirror layers. The delocalization of light propagating in modes within the bandgap and localization of light in modes outside the band gap was explored in detail in the context of Anderson localization in Ref. .
Soon after the first successful melt-processed DBR laser, the same research group blended laser dyes directly into one of the two extruded polymers, producing a true DFB multilayer structure. These first co-extruded roll-to-roll processed DFB lasers were fabricated from dye-doped poly(styrene-co-acrylonitrile) with 25 wt % acrylonitrile (SAN25) alternated with layers of a fluorelastomer terpolymer of vinylidene fluoride, hexafluoroproplyene tetrafluoroethylene (Dyneon THV 220G) (THV), and skin layers of low density polyethylene (Dow LDPE 6201). The refractive indices of the two constituent polymers at 633 nm were found to be 1.57 and 1.37, respectively, producing highly reflective films from 64 and 128 layer stacks. The laser dyes used were commercial R6G and a newly synthesized 1,4-bis-(α-cyano-4-methoxystyryl)−2,5-dimethoxy-benzene (C1RG, absorption max at 434 nm, fluorescence peak at 515 nm) which were then pumped by 7 ns, 10 Hz, Nd:YAG based optical parametric oscillator at 532 nm and 489 nm, respectively. Both dyes were insoluble in the lower index THV polymer and so were solution blended in chloroform into SAN25 and then further diluted with neat SAN25 into a nominal 1 wt % dye concentration. The THV/SAN25 pairing was chosen because THV acted, in this case, as a barrier layer for the dye, sequestering the dye in the SAN25 during melt processing. Though the melt processing temperatures are not so high as to destroy the laser dyes, the dyes tend to diffuse through many polymers, even to the point of leaving the multilayer system during extrusion. The effectiveness of the THV as a barrier layer appears to depend upon the polarity of the dye. It provided an effective barrier for R6G, but some diffusion was still evident with the C1RG dye.
Although 128 layer films exhibited the desired reflectivities in the bandgap region, it was found that stacking films with 64 layers improved the uniformity and performance of the system. The highest slope efficiency of about 8% was seen in a stack of five 64-layer films, which showed a lasing threshold of 100 µJ/cm2. As expected for the complicated band structure that resulted from nonuniform layer thicknesses, lasing occurred at a defect state within the bandgap, rather than at the long wavelength bandedge. Spectra of the multilayer films depend on the relationship between the reflection band, cavity interference, dye absorption and re-absorption. The reflection band not only shifts to the blue with increasing incidence angle, but the shape of the band changes as well due to the relative contribution of dye absorption across the band. As expected, the lowest thresholds appeared when the pump angle of incidence was at a transmission maximum, entirely outside of the reflection band.
The coherence length of these lasers has not been carefully measured, but a crude estimate follows from the observation of the pronounced etaloning rings shown previously in Figure 8(b). These rings indicate a coherence length that is at least several film thicknesses. This estimate is consistent with the group velocity minimum at the interband defect, where the lasing occurs. As one can see from Figure 3(c), it is not unreasonable to expect residency times in the cavity that are 10 to 20 times the naive transit time, assuming perfect layer/film uniformity. Based on simulations that include layer non-uniformity at the level measured for actual multilayer films [see Fig. 10(a)], the expected coherence length of the laser output must be greater than a few multiples of the film thickness. Additionally, a crude upper bound on the coherence length is afforded by the experimental observation that the laser output closely mirrors the pump pulse shape, indicating the coherence length is smaller than 20 cm.
With the naive laboratory estimated threshold fluence of roughly 10 kW/cm2, coextruded DFB lasers clearly await significant improvement to achieve the commercial goal of being LED or LD pumped. Thankfully, because much of the work described herein represents just the first attempts with these nano-structured plastic lasing materials, there are several “low hanging fruit” that lead immediately to lower lasing thresholds. Improvements in layer uniformity could bring nearly an order of magnitude improvement in cavity Q, as could concomitant optimization of dye characteristics and dye density within the structure. As previously discussed, simply folding the DFB on the low index component laden with gain media will lead to immediate lowering of the threshold.
Further improvements such as surface patterning to reduce the cavity Q of any parasitic transverse modes also appear straightforward, as previously described. Looking ahead, we have experimentally verified material compatibility by bonding these films directly to laser diode output facets without impairing the lifetime of the diode or degrading the film's optical properties. Of course, dye lifetime, as described elsewhere in this review, may become a serious hurdle in the way of commercial utility of plastic DFB's, but lowering the lasing threshold is expected to somewhat ameliorate that issue.
TUNABILITY IN MULTILAYER POLYMER DFBS
One of the historic strengths of organic lasers is that they can be widely tuned by using the broad emission spectra of the gain media and the tight dependence of the lasing wavelength on the optical structure of the resonator cavity. For conventional liquid and Fabry-Perot cavity dye lasers, tuning is usually accomplished by a separate rotating reflective diffraction grating element in the resonator.[160, 161] For multilayer DFB lasers, tunability is enabled through internal manipulation of the multilayer resonator itself. In some cases, this leads to even broader tunability than seen with the same dyes in conventional Fabry-Perot resonators. For example, with R6G dye, the typical liquid solution dye laser is tunable over a 40 to 50 nm range, but with the same dye in multilayer polymer films with different preferred defect lasing modes, we have observed lasing over a range from 560 nm to 650nm.
Three independent tuning mechanisms have been demonstrated with co-extruded polymer DFB lasers. These are (i) structure design tuning through the use of a terraced defect center layer, (ii) mechanical tuning through the use of elastomeric polymers, and (iii) temperature tuning using the thermal expansion characteristics of the constituent polymers. While none of these modalities, taken singly, is previously unrecognized, we believe their simultaneous employment in multilayer polymer DFB structures is unique and likely advantageous for applications.
As a point of comparison, note that corrugated grating DFB lasers have also been tuned by refractive index changes,[165, 166] stretching,[167, 168] thermo-refractive effects, and defect layer thickness changes.[170, 171] Grating-based DFB lasers in waveguide structures also have been shown to be tunable by varying the thickness of the film over the grating, independently from the grating spacing. Discrete spectral tuning and switching using defects has also been demonstrated in cholesteric liquid crystal DFB systems.[173, 174]
As previously described, one of the possible ways to improve upon a simple multilayer DFB structure is to insert a phase slip defect, e.g., a layer towards the middle of the stack that breaks the periodicity and creates a corresponding spectral defect in the reflection band. Lasing preferentially occurs at the spectral defect location in the reflection band and, by modulating the thickness of defect layer, one can continuously tune the laser output wavelength.
Even in the presence of random variations in the layer thickness around the ideal quarter-wave stack, intentional phase-slip defects can be used to improve and control lasing. Use of deliberate phase-slip defects to control lasing was demonstrated in co-extruded multilayer R6G-doped DFB laser films. Lacking a substrate, these melt-pressed films (comprised of alternating SAN25 and THV layers, as described above) are particularly amenable to simple postprocess folding to create a phase-slip half-wavelength defect. In this case, folding versus stacking of the multilayers led to a factor of 3 to 6 times improvement in lasing efficiency and a lower lasing threshold (see Fig. 11).
Further, by using a weak solvent to terrace the layer that would become the centerfold, structure defects of stepped thickness were fabricated, as shown in the right most schematic of Figure 1(c). This enabled tuning of the laser in discrete steps. By simply shifting the lasing spot on the film a matter of millimeters, and taking advantage of the broad fluorescence of R6G dye, the wavelength was tuned across much of the reflection band (see Fig. 12). Sensitivity of the lasing wavelength to variations in layer thicknesses is discussed more extensively in Ref. .
As the center layer is thinned, the spectral defect moves to shorter wavelengths until it disappears into the band edge when the thickness corresponds to the standard quarter-wavelength optical thickness. When the center layer is thinned still further, a spectral defect in the bandgap will reappear at the long wavelength reflection band edge and again shift to shorter wavelengths. Of course, if the layer is thinned to vanishing, the defect state again appears as a center fold defect, but this time due to a doubling of the thickness of the other refractive index constituent now at the center.
The spectral location of the bandgap can be made mechanically tunable by fabricating the multilayer laser film using elastomeric polymers. While this feature has not yet been realized fully in a multilayer DFB structure, it has been demonstrated in DBR films, first by spin coating, and, more recently, by co-extrusion of elastomers.[154, 163] In the latter case, 128 alternating layers of THV, and ethylene-octene (EO, n ≈ 1.48) were co-extruded with a R6G dye-doped Lotader elastomeric skin layer, which acts as the thick active central gain medium when the films are folded to create the microcavity DBR laser. The Lotader skin layer is a blend of ethylene terpolymer with 40% by weight acrylic ester and glycidyl methacrylate [Arkema, www.lotader.com]. The average layer thickness was approximately 110 nm and the center layer thickness was approximately 30 µm after folding.
A pseudo-affine model relates the shift λ/λ0 of the reflection band center wavelength, λ0, as a function of the lateral linear strain, Δl/l0,
where are the initial and the final indices of refraction in the direction of stretching and perpendicular to the direction of stretching, respectively. For small strains (<0.2), the induced birefringence was almost negligible , yet stretching the laser film yielded nearly continuous wavelength tuning from red to green (from ∼625 nm to ∼570 nm), as seen in Figure 13.
Whether laser output variability with temperature is desirable depends upon your point of view. Many applications require a stable output frequency, and drifts due to temperature must be carefully suppressed. Other applications, however, benefit greatly from the ability to tune the wavelength simply by changing the temperature. Fortunately, due to the wide range of polymer choices, it is possible to design a multilayer polymer DFB system for either type of application. The key, however, is to consider not only the thermo-mechanical and optical properties of the polymer constituents, but also how those constituents will interact when constrained in the multilayer.
Most, but not all, polymers expand with temperature, which leads to two competing effects on a binary multilayer structure, the layer thicknesses increase, but the indices of refraction drop. Because the optical pathlength is the product of the thickness and refractive index, these two effects partially offset one another, but do not, in general, cancel. The situation in a multilayer is further complicated, however, when the two (or more) constituent materials want to expand at different rates. This nonuniform expansion is problematic in a multilayer because the layered materials are tightly fused together (in many cases involving interdiffusion regions of tens of nanometers or more) and are thus constrained to maintain the same expansion rates along the interfaces. (It is assumed, and seen experimentally, that they do not delaminate.)
In a bilayer system, the difference in thermal expansivities leads to the bending of one material over the other, as in the familiar bimetallic strip, but this is not possible in a system of tens or even hundreds of layers. Instead, if the two polymers are not matched in their expansion rates, the rate along the interfaces will be dominated by the material that is more rigid (typically, but not necessarily, the polymer with the smaller thermal expansivity). In compliance with Poisson's ratio relating the bulk expansion rate to the linear expansion rate, either the lower expansivity layers will not thicken as much as expected perpendicular to the film plane because they are forced to expand more in the directions along the interfacial plane, or the polymer layers having higher expansivity will thicken more than otherwise expected as they are unable to expand along the interfacial plane. Measuring, let alone predicting, the response given a particular pairing of polymers in a multilayer is an ongoing area of research. To measure the changes in thickness and refractive index directly requires a level of precision that is easily obtained by interferometry.
Fortunately, observation of changes with temperature in the optical band structure and laser output, together with the multilayer modeling methods described above, greatly aid in our understanding of these effects. We have explored the thermo-optic spectra of both polymer DRB and DFB laser structures. The DFB systems were comprised of alternating layers THV and SAN25 as previously described, which were folded alternately onto the THV or SAN25 layer. In this system, modeling of the measured response indicates that the more rigid SAN25 polymer constrains the in-plane expansion of the THV polymer, which responds in the multilayer by increasing in thickness by nearly a factor of three more than would be expected from isotropic properties. As the temperature of the system increased, the reflection band not only shifted to longer wavelengths, it also expanded, with the long wavelength band edge shifting more than the short wavelength band edge. Due to layer thickness variations throughout the stack and the field energy dependence of the center layer defect, lasing appeared near opposite band edges depending upon polymer in the center defect layer. The spectral location of the lasing wavelength at opposite band edges leads to different thermo-spectral coefficients for the two different folded DFB lasers, as can be seen in Figure 14.
Note that in each case, the broad emission spectrum of the dye also changes slightly with temperature, but this change has little effect on the lasing wavelength which is otherwise dominated by the dispersion of the DFB resonator cavity and the location of the dye layers.
To explore further the thermal effects of combining different polymers, we also studied two very different DBR systems, each with the same R6G gain layer. These results, though not from a DFB system, are important because they are directly relevant to the design of a DFB laser for a desired thermal response. Briefly, when the DBR mirrors were made of alternating layers of THV and Ethylene Octene (EO) polymers, the mismatch in thermal expansivity between EO and THV led to a large thermo-spectral response on the order of a third of a nanometer per °C temperature increase. However, when the same gain layer was placed between DBR mirrors comprised of the polymers PMMA and PS, whose thermo-elastic properties were well matched, the system showed an order of magnitude less sensitivity to temperature, ∼0.03 nm per °C (see Fig. 15). Thus, the pairing of polymers in the DFB multilayer has a significant impact on the laser wavelength temperature dependence. Studying these temperature dependent spectral changes also provides new insights into how the two polymers respond to being constrained together in the multilayered system. The study of postextrusion constraints on multilayer systems is a rich area for further research, with possible device/sensor applications beyond lasers.
PROSPECTS AND PROMISES
We have demonstrated that co-extruded multilayer polymers promise to be an easily mass-produced laser material offering advantages over corrugated grating-based DFB lasers due to the large area available for lasing, flexibility and ease of manipulation, and wide range of mechanisms available for tuning the output. Further work will be needed to achieve some of the performance characteristics that have been demonstrated in long-studied corrugated grating-based DFB lasers such as low thresholds, extended operation, and the use of microchip lasers or even LEDs as pump sources. Further improvements in tunability and extended operation are possible with improved dyes and/or quantum dot emissive sources.[178, 179]
In order to achieve lasing with low cost and/or continuous wave pump sources, lasing thresholds will need to be lowered by more than an order of magnitude from the earliest coextruded systems. Improved layer uniformity, increased refractive index differences, lower parasitic absorptive and scattering losses, better dyes and other emissive species, and use of optimized defect structures suggest that performance parity with other DFB lasers is within reach.
Indeed the use of inexpensive optical pump sources, such as laser diodes, diode-pumped solid state lasers (DSSLs) or even light emitting diodes may be the method for indirectly achieving the elusive “holy grail” of electrically pumped widely tunable polymer lasers. While, obviously a different approach than direct electrical pumping of organic semiconductor lasers,[121, 180, 181] this could possibly fill a similar niche. Even while efforts are underway to bring the thresholds down, progress is also being made to improve upon possible pump sources.
Low cost, small, tunable laser sources are applicable to specialty spectroscopic systems, such as chemical sensors, and even DNA sequence sensors. Dye lasers continue to be widely used in medical applications, and, for example, low-cost tunable polymer DFB lasing surfaces could be used even as disposable laser sources for dermatological applications applying the multilayer laser films directly to the skin. In light of the technical developments outlined in this review and the great portent of manifold economical applications, further sustained development efforts towards this versatile class of laser materials promise to reap widespread benefits.
The authors are grateful to the National Science Foundation for financial support from the Science and Technology Center for Layered Polymeric Systems under grant number DMR 0423914 and to the State of Ohio, Department of Development, State of Ohio, Chancellor of the Board of Regents and Third Frontier Commission, which provided funding in support of the Research Cluster on Surfaces in Advanced Materials. The authors also thank Dr. Nathan Dawson and Michael Baker for research assistance.
James H. Andrews received his Ph.D. in Physics from Case Western Reserve University (Case) (1995) while researching organic nonlinear optical materials. Dr. Andrews joined the faculty at Youngstown State University (YSU) in 1996 where he is currently a professor. His research is primarily on coherent optical processes, particularly in structured materials.
Michael Crescimanno received his Ph.D. in Physics from University of California, Berkeley (1991) for various studies in low-dimensional quantum field theories, gravitation and string theory. His most recent work is in the areas of quantum optics, optics and mathematical physics. Dr. Crescimanno joined the physics faculty at Youngstown State University in 2000 where he is currently a professor.
Kenneth D. Singer received his Ph.D. in Physics from the University of Pennsylvania (1981) for studies of nonlinear optics in organic materials. Following 8 years as a member of the technical staff at Bell Laboratories, he joined the faculty at Case Western in 1990. His research centers on optical and electronic properties of organic materials. Dr. Singer is currently the Ambrose Swasey Professor of Physics at Case Western Reserve University.
Eric Baer received his degree of Doctor of Engineering from Johns Hopkins University in 1957 while researching heat transfer in condensation. His most recent work is in the areas of micro- and nanolayered film systems and applying lessons from nature to the development of polymeric systems. Professor Baer joined the faculty of Case Western Reserve University in 1962. He is currently The Distinguished University Professor and Director of the NSF Science and Technology Center on Layered Polymeric Systems.