1. Introduction

Lithography, from Ancient Greek λίθος (lithos) meaning stone, and γράφειν (grafein), meaning ‘to write,’ is a printing technique that originally involved drawing on a polished slab of limestone with either grease or a liquid called tusche. Photolithography is a modern form of lithography in which exposure to light is the means of drawing. Photolithography has played a dominant role in the fabrication and mass production of integrated circuits since the 1960s. The advance of photolithographic techniques has guided the revolution in the microelectronics industry. The demand for ever-decreasing dimensions of transistors, as embodied by Moore’s law, typically necessitates photolithographic exposure at the shortest possible wavelength [1]. At the same time, emergent nanotechnologies require the miniaturization of mechanical, chemical or medical systems for lab-on-a-chip applications in areas such as photonics, biophysics, and optofluidics. These applications, however, often require the development of new lithographic tools, as there are some important limitations to the application of conventional photolithography in the fabrication of micro- and nano-devices.

One such constraint is that cutting-edge conventional photolithography requires the use of expensive masks and custom-made optical components, which makes the technique viable only for large-scale manufacturing [2]. Another issue is that the photoresists that are formulated for photolithography require severe post-processing conditions. Last but not least is the fact that conventional photolithography is in principle a planar technique, as it only allows two-dimensional fabrication at the surface of the resist. This issue can be addressed using layer-by layer fabrication, but control over the third dimension is still limited compared to the other two.

To overcome these issues, efforts have been focused on developing alternative lithographic approaches that can operate in 2.5 or 3 dimensions [3]. Such approaches include self-assembly [4], nanoimprint lithography [5], microstereolithography [6], and 3D ink-based printing [7]. Each of these techniques has its own inherent limitations, which may include requiring a large number of processing steps, creating structures that may have poor mechanical properties, or offering a maximum resolution of a few microns. Another class of alternative approaches is laser-based techniques, such as multibeam interference lithography [8] and phase-mask lithography [9]. These techniques can offer sub-micron resolution, but a major drawback is that because these methods are based on light diffraction and interference, only periodic patterns can be created.

A powerful tool for truly 3D fabrication is multiphoton absorption polymerization (MAP) lithography [10]. MAP has emerged as a versatile means for mask-free fabrication of fully 3D structures with a transverse feature size of 100 nm or less, using near-infrared excitation (typically at 800 nm) [11–13]. In MAP, multiphoton absorption is used to expose a photoresist one volume element (voxel) at a time. The use of high-repetition-rate fs oscillators along with high-precision mechanical stages for MAP allows for the creation of intricate, arbitrary 3D shapes over dimensions of hundreds to thousands of microns or more.

In MAP, an electronic transition in a material is driven by two or more photons in combination. The probability for the simultaneous absorption of n photons via laser excitation of a molecule is proportional to the n-th power of the intensity of the light employed. Owing to this nonlinear intensity dependence, by tightly focusing a laser beam by means of a microscope objective it is possible to achieve the high peak intensity needed to drive the multiphoton process only within the minute focal volume. On one hand, this nonlinearity limits the exposure to the focal volume. On the other hand, unwanted exposure is prevented from occurring along the beam propagation axis, allowing for true 3D fabrication.

The narrowing of the point spread function (PSF) for exposure due to optical nonlinearity is not enough to explain the small feature sizes attainable with MAP. In negative-tone photoresists, there is an exposure threshold below which quenching dominates and prevents crosslinking. MAP is typically performed using beams with a Gaussian spatial profile. Therefore, operation at appropriate exposure conditions can constrain the polymerization to occur only in the most intense region of the Gaussian beam.

Despite the subdiffraction feature size in the transverse plane, the axial feature size is in general considerably larger. As a result, MAP voxels are ellipsoids, with the axial dimension being 3-5 times greater than the transverse one, as it is dictated by the physics of the focusing of a laser beam through a microscope objective and the concomitant non-symmetrical wavefront that is formed [14]. This issue poses major challenges for the fabrication of high-resolution structures that require uniform features that are of the minimum possible size in all dimensions [15]. Another important facet of multiphoton lithography is that it is essentially a far-field optical technique. Thus, as in conventional optical imaging and photolithography, MAP is limited by diffraction.

After the demonstration of subdiffraction resolution in fluorescence microscopy using two-color approaches, it was recognized that a similar concept could be implemented in photolithography. This realization inspired the creation of new photoresists that respond differently to two colors and/or two beams of light. In a typical implementation of this concept in optical lithography, one laser beam is used for linear or two-photon exposure, and a second laser beam is used to prevent the effects of the exposure. By appropriate spatial shaping of these two beams, it is possible to improve both the feature size and resolution of optical lithography. This paper reviews the past decade of advances in two-color approaches to resolution enhancement in linear and multiphoton lithography.

We begin this review with a discussion of the optical technique that inspired the establishment of the field of two-color lithography (2CL), stimulated emission depletion (STED) fluorescence microscopy. We then describe the 2CL approaches that have been demonstrated to date, and discuss the state of the art for each method. We conclude with a discussion of the challenges that must be addressed in the near feature to push the limits of two-color lithographic techniques.

2. Stimulated emission depletion microscopy

Fluorescence microscopy has historically faced challenges similar to those of photolithography. The most important drawback is that optical imaging and photolithography are essentially far-field optical techniques, and as a consequence are limited by diffraction. In imaging, the resolution refers to the shortest distance between two points on a specimen that can be distinguished by the observer or camera system as separate entities. Similarly, in photolithography resolution refers to the shortest repeat distance between two printable features. This limitation imposed by the wave nature of light is usually expressed by the Abbe criterion:

In the mid-1990s Stefan Hell suggested the use of far-field optics to overcome the diffraction limit resolution in optical imaging, through the concept of stimulated emission depletion (STED) microscopy [18]. STED was the first of a class of superresolved techniques called reversible saturable optical fluorescence transitions (RESOLFT) that use molecules that have at least two distinguishable states between which reversible switching is possible, and in which at least one such transition can be optically induced [19]. A decade later, the family of super-resolved techniques for optical microscopy expanded, as it became apparent that the use of fluorescent proteins that display a controllable photochromism, such as green fluorescent protein, can be used to prevent nearby fluorophores from being excited synchronously [20,21]. The significance of the research in this field is reflected on the fact that methods that achieve superresolution were the subject of the 2014 Nobel Prize in chemistry.

The STED principle is based on the fact that whereas the fluorescence is emitted isotopically, stimulated emission travels in the direction of the stimulating wave. Thus, in a STED microscope, only the fluorescence emission is detected. A laser pulse at wavelength λ_exc excites the fluorophores, and a second laser pulse at λ_SE stimulates emission (Fig. 1(a)). If the second pulse arrives before the molecules have had time to emit, the excited states will be depleted via stimulated emission. If the PSFs of the excitation and deactivation beams overlap completely, then there is no improvement in resolution. In STED microscopy the excitation beam is focused to a standard PSF. However, the deactivation beam passes through a phase optic before it is focused, causing it to have a dark region in the center of its PSF (Fig. 1(b)). In the dark region of the deactivation PSF, excited molecules do not undergo stimulated emission. This scheme results in confinement of the region from which fluorescence is detected. The nature of the phase optic determines whether the improvement in resolution is in the beam propagation direction or transverse to this direction. Furthermore, two separate deactivation beams that pass through different phase optics can be used to improve the axial and transverse resolution simultaneously [22,23]. Obviously, the dimensions of the dark region of the deactivation PSF are limited by diffraction. However, operating at a deactivation intensity that is in the saturation regime can cause the effective dimensions of the bright region of the PSF to be enlarged, thus allowing the fluorescence area to become arbitrarily small in principle.

 

Fig. 1. (a). Jablonski diagram showing excitation at λ_exc, red-shifted stimulated emission (λ_SE), and fluorescence. (b) Excitation spot (xy plane, left), doughnut-shape deactivation spot (center) and the effective PSF for fluorescence (right).

Download Full Size | PDF

The resolution that can be attained in this context can be formulated by rewriting the Abbe criterion as [18,19]

To gain more insight into the molecular mechanisms of STED so that it can be applied to photolithography, it is important to understand the underlying photophysics and to identify the possible pathways that the excitation can follow. A schematic diagram of the photophysics of STED fluorescence microscopy is shown in Fig. 2. First, the excitation pulse promotes the fluorophores from the ground electronic state (S₀) to an excited vibrational level of its first excited electronic singlet state (S₁). From there, the molecules relax rapidly (∼ few ps) to the vibrationless level of S₁ via intramolecular vibrational redistribution (IVR) or intermolecular collisions. We assume that the fluorescence lifetime, τ_fl, is much longer than IVR timescale, so that the majority of the fluorescence occurs from the vibrationless level of S₁. At this point, a second pulse can switch the molecules off by forcing them down to the manifold of vibrational levels of S₀. For this process to occur, the stimulated emission rate σI_SEλ_SE/hc must be much larger than that of the spontaneous decay,1/τ_fl, where σ is the cross-section for stimulated emission, h is Planck’s constant, c is the speed of light in vacuum, and I_SE is the deactivation beam intensity. This vibrationally-excited electronic ground state loses energy via IVR, depopulating the lower level in the stimulated-emission transition. We assume that excitation and emission transitions are vertical, i.e. that the nuclear coordinates do not change during the electronic transitions. The minimum of the S₁ potential surface is usually shifted with respect to the S₀ surface, resulting in a red-shift of the emission transition compared to the excitation transition. In general, a large Stokes shift is desirable in fluorescence microscopy, as this condition prevents reabsorption.

 

Fig. 2. Schematic representation of the photophysics of STED fluorescence microscopy. The molecule is excited from the ground vibrational state of the ground electronic state S₀ to a vibrational state of the first excited electronic state S₁. The excited molecule relaxes to its ground vibrational state via intramolecular vibrational redistribution (IVR). From there, it can fluoresce or undergo stimulated emission to a vibrationally excited state of S₀. Stimulated emission also competes with excited state absorption (ESA) from S₁ to higher electronic states.

Download Full Size | PDF

A typical fluorophore has fluorescence lifetime on the order of nanoseconds or longer, providing ample time for deactivation. The use of deactivation pulses with durations on the order of tens of picoseconds allows for efficient stimulated emission. The deactivation pulse should also be substantially longer than the IVR time in the ground electronic state to avoid saturation of the deactivation transition.

Another important aspect that needs to be considered is the potential for disruption of the STED process by competing processes. For instance, the deactivation beam also has the potential to promote the fluorophore to a higher excited state. This process is known as excited state absorption (ESA). In the presence of ESA, the molecules promoted to a higher excited state will no longer be available for STED, and therefore the deactivation efficiency drops. It is also difficult to predict the relaxation channel that such highly excited molecules will follow. Some will return back to S₁, but others may undergo irreversible photochemistry or intersystem crossing (ISC).

3. Two-color approaches in lithography

The success of STED microscopy inspired the use of 2CL approaches to enhance the resolution of visible-light photolithography. In 2009, three groups demonstrated sub-diffraction nanopattering using more than one light source [15,25,26]. In these approaches, one optical beam controls the spatial distribution of exposure and another beam triggers chemical activation. Although the inspiration for 2CL stemmed from STED microscopy, none of the first three groups that demonstrated 2CL used this mechanism as a means for deactivation. This fact underscores the complexity of the photophysics that exists in a photoresist.

Over the last decade, a number of groups around the world have been working on 2CL. The deactivation mechanisms in these approaches include stimulated-emission depletion, photoinduced electron back-transfer, photoinduced radical inhibition, photoisomerization, and triplet absorption. Here we review the most significant approaches that have used two-color approaches for enhancement of lithographic resolution.

3.1 Stimulated emission depletion

In principle, STED should be able to enhance lithographic resolution in negative-tone photoresists. In close analogy to a STED fluorophore, an ideal photoinitiator should leave enough time between excitation and bond cleavage to allow for efficient deactivation. This condition seems to be met in most Norrish Type I photoinitiators, which form radicals after undergoing ISC from a singlet state to a triplet state (Fig. 3). Typical timescales for ISC are on the order of 100 ps [27]. Although this time window is significantly less than the fluorescence lifetime of a typical fluorophore (∼ ns), it should be long enough to be able to drive the molecule back to the ground state through stimulated emission before ISC occurs. This timescale favors the use of lasers with pulse durations in the ps regime.

 

Fig. 3. Schematic representation of the photophysics of STED in MAP. The photosensitizer begins in the ground vibrational state of the electronic ground state S₀, and is driven to a vibrationally excited state of the first excited electronic state S₁. The electronically-excited molecule undergoes IVR to the bottom of the S₁ manifold. From there, the molecule can undergo ISC to the chemically active triplet state T₁, or undergo stimulated emission to the manifold of vibrational levels of S₀. The system can be also driven through ESA to a higher electronic state by the deactivation beam, which tends to be more reactive.

Download Full Size | PDF

We have investigated the possibility of deactivation via STED in a wide range of Type I radical photoinitiators that are used widely in MAP [28,29], such as 2,4,6-trimethylbenzoylphenyl phosphinate (Lucirin TPO-L) and 2-benzyl-2(dimethylamino)-1-[4-(morpholinyl) phenyl)]-1-butanone (Irgacure 369). Surprisingly, a laser tuned within the fluorescence spectrum of these photoinitiators does not result in polymerization inhibition, but rather increases the rate of polymerization. These findings indicate that the deactivation beam causes ESA more efficiently than STED. The rate of polymerization in this case is enhanced, most likely because the higher excited state is more reactive due to faster ISC than from the first excited singlet state. The difference in the ability to deactivate via stimulated emission between photoinitiators and dyes stems from inherent differences in the photophysics of these classes of materials. The optical excitation of typical photoinitiators is governed by relatively weak n–π* transitions with low oscillator strength. On the other hand, typical fluorescent dyes are governed by aromatic π–π* transitions with large extinction coefficients, and therefore strong fluorescence. The oscillator strength for absorption between two states is similar to the oscillator strength for stimulated emission. Therefore, a molecule with large extinction coefficient between S₀ and S₁, such as a dye, is expected to exhibit strong stimulated emission from S₁ to S₀ but weak ESA from S₁. The opposite holds true for a molecule with small extinction coefficient, such as a photoinitiator, which is expected to exhibit weak stimulated emission but strong ESA.

The above discussion makes apparent that to use STED as a means for polymerization inhibition in 2CL, it is essential to identify photoinitiators that have large extinction coefficients. These criteria were found to be met in the laser dye 7-diethylamino-3-thenoylcoumarin (DETC), which can act as an effective multiphoton photoinitiator and exhibits polymerization inhibition with 532 nm light. This system allowed Wegener and coworkers to attain significant axial and transverse feature size improvement as compared to the same photoresist used without the deactivation beam [30]. Wollhofen et al. later used a photoresist containing DETC to achieve feature sizes of 55 nm at a 120 nm pitch using pulsed, 780 nm, multiphoton excitation and CW, 532 nm deactivation [31].

STED has some advantages as a method for deactivation. Stimulated emission can follow rapidly after excitation, given the large oscillator strengths of typical dyes. Immediately after stimulated emission and the subsequent vibrational relaxation the photoinitiator molecules are available for re-excitation. In this way, scanning speeds in the range of cm/s or even higher might be achieved for point-to-point fabrication. Furthermore, in STED, the photoinitiator molecules return back to their ground electronic state, which is desired because in this way unintended transfer to a reactive state can be avoided.

However, the above advantages do not mean that STED is necessarily an ideal mechanism for deactivation in 2CL. A kinetic model of the dynamics of STED deactivation in MAP makes it evident that this approach has some disadvantages [32]. First, the time available for deactivation is limited by the apparent fluorescence lifetime. Furthermore, after excitation some time is required for IVR in the excited electronic state before stimulated emission can take place. Stimulated emission will bring the excitation to a higher vibrational level of the ground electronic state, so there is some time required for IVR in that state as well. Therefore, the time window for stimulated emission will be after IVR in the excited state but it is limited by the rate of the IVR in the ground state, and so is typically in the range of a few picoseconds to few hundreds of picoseconds. This fact makes the use of CW deactivation inefficient, as the majority of the deactivation photons remain unused.

To be compatible with the IVR and ISC time scales, deactivation via STED is generally best performed using synchronized pulses with durations on the order of few picoseconds to tens of picoseconds. In this case the deactivation can be efficient, but there is a limit for the maximum power that can be employed, because at high excitation intensities unwanted phenomena such as ESA and/or saturation of deactivation can occur. Also, the high peak intensity of pulsed irradiation leads to nonlinear absorption.

3.2 Photo-induced electron back-transfer

The concept of resolution augmentation through photo-induced deactivation (RAPID) lithography was introduced in 2009 [15]. RAPID was the first 2CL technique that used multiphoton absorption for excitation. In the original demonstration, malachite green carbinol base (MGCB) was used to initiate radical polymerization via two-photon absorption of 800 nm light in an acrylic negative-tone photoresist. The laser pulses had a duration of ∼150 fs. The addition of a second 800 nm laser beam resulted in inhibition of polymerization. Experiments were conducted at different delay times between the excitation and deactivation laser pulses, up to maximum delay of 13 ns, which corresponds to the repetition time of the Ti:sapphire lasers that were used. Interestingly, it was found that the efficiency of deactivation was independent of the delay time over this time window. This finding was surprising, because it indicates that the lifetime of the state that is deactivated is much longer than 13 ns. The fluorescence lifetime of a fluorophore with a large absorption cross section is expected to be no more than a few ns, so the deactivation of MGCB must take place via a mechanism that does not involve stimulated emission. Further experiments revealed that the polymerization could be inhibited completely even by using CW laser irradiation, as it is evidenced in the SEM image in Fig. 4(a). However, in this case, there is a certain CW laser power above which the polymerization cannot be inhibited. This observation indicates that there are two competing channels that can lead to polymerization, only one of which can be photoinhibited. If the polymerization channel that cannot be inhibited generates enough radicals to overcome the threshold for polymerization, then the deactivation beam cannot prevent the polymerization.

 

Fig. 4. (a) Two-color RAPID lithography. Initiation of polymerization is achieved via an ultrafast laser at 800 nm, whereas the inhibiting source is an 800 nm CW laser beam that is chopped. (b) Proposed deactivation mechanism in RAPID. Following excitation, MGCB ejects an electron, which is solvated in the photoresist. If the solvated electron remains in proximity with the cation, then photoinduced back-transfer can regenerate the photoinitiator. If the electron diffuses away from the cation, then recombination with the cation can no longer take place. (c) Voxel height and aspect ratio as a function of the deactivation power. (d) The upper panels show 3D and contour AFM images of the minimum feature (voxel) created using RAPID lithography. The lower panels show the corresponding images of the smallest voxel created with conventional MAP lithography. (a), (c) and (d) are reproduced with permission from ref. [15], copyright American Association for the Advancement of Science.

Download Full Size | PDF

The proposed mechanism for the initiation and photoinhibition of polymerization with MGCB is based on solvated electrons (Fig. 4(b)). Initially, the excitation of MGCB leads to electron ejection. The electron is solvated in the surrounding photoresist. If the electron remains in proximity to the MGCB cation, subsequent irradiation via the deactivation beam can lead to back-transfer, regenerating the photoinitiator. This mechanism is consistent with a number of experimental observations. First, if the solvated electron diffuses away from the MGCB cation the polymerization cannot be inhibited, which can explain the existence of the second channel. This picture is in line with the fact that the increase of the photoresist viscosity makes the deactivatable channel more dominant. Additionally, the relatively low reactivity of the solvated electron is consistent with the fact that initiation of polymerization is found to be slow enough to be inhibited even with a CW laser. Further experiments on other standard dyes indicated that the solvated electron mechanism may be quite common for initiation of polymerization [33]. As shown in Figs. 4(c) and (d), RAPID with the 800 nm pulsed excitation and a phase-masked 800 nm CW deactivation enabled fabrication of features with scalable axial dimension, down to a 40-nm minimum feature size. 40 nm corresponds to 1/20 of the excitation and deactivation wavelengths. RAPID demonstrated unambiguously the ability to use 2CL to fabricate 2D and 3D structures with a feature size that is a small fraction of the wavelength of the light employed.

One of the advantages of RAPID is that the state that is deactivated has a long lifetime. Deactivating from a long-lived state allows for the use of CW radiation of low intensity. The use of low deactivation intensity is desirable because it prevents unwanted phenomena such as multiphoton excitation or thermal effects from occurring. The long lifetime of the deactivatable state also helps to minimize the competition between inhibition with initiation.

On the other hand, it should also be noted that the long time between excitation and the initiation of cross-linking can limit the velocity that can be attained for point-by-point fabrication. As in STED fluorescence microscopy, the use of the shortest possible wavelength for deactivation is expected to give the best resolution in RAPID. For MGCB, the deactivation efficiency drops drastically at wavelengths shorter than 800 nm. As noted above, there is a limit to the deactivation intensity that can be used, because there is a polymerization channel that cannot be deactivated when MGCB is used as photoinitiator.

3.3 Photoinhibitors

Another approach to 2CL is based on photoinhibitors. A photoinhibitor is a molecule that can act as a radical trap upon excitation, and therefore can lead to polymerization quenching. McLeod and coworkers developed a 2CL irradiation scheme based on photoinhibition [26]. CW 473 nm light was used to excite the photoinitiator species in a methacrylic photoresist, and CW 364 nm light was used to excite the inhibiting species. The deactivation beam was phase shaped in a Gauss-Laguerre “doughnut” mode surrounding the focal point of the primary beam used to initiate polymerization. This approach enabled the creation of features as small as 64 nm in diameter, even though the excitation was based on linear absorption. Gu and coworkers further demonstrated the creation of 40 nm features using a modified resin that was based on the same photoinhibitor [34]. In 2013, Gu and coworkers also developed a photoresist that used multiphoton excitation and linear deactivation via photoinhibition [35]. They reported a minimum feature size of 9 nm, and were able to create a pair of lines of 20 nm width in a pitch of 52 nm.

Recently, de Beer et al. demonstrated a method for rapid and continuous stereolithographic additive manufacturing by using two-color irradiation of (meth)acrylate resins containing complementary photoinitiator and photoinhibitor species [36]. The resins were made using camphorquinone (CQ) and ethyl 4-(dimethylamino)benzoate (EDAB) as a visible light photoinitiator and co-initiator, respectively, and bis[2-(ochlorophenyl)-4,5-diphenylimidazole] (o-Cl-HABI) as a photoinhibitor. o-Cl-HABI shows weak absorbance in the blue region of the spectrum and moderate absorbance in the near ultraviolet (UV). In contrast, CQ absorbs blue light (λ_max = 470 nm) strongly, whereas it absorbs poorly in the near UV. The minimal overlap in the absorbance spectra of CQ and o-Cl-HABI in the near UV to blue region of the spectrum enabled polymerization to be initiated selectively with blue light and inhibited with UV light.

The use of photoinhibitors for deactivation in 2CL has both attractive features and inherent disadvantages. One favorable characteristic is that the photoinitiation of polymerization is decoupled from the deactivation process, i.e., the molecules that create radicals are not the same as those that inhibit polymerization. However, a drawback is that both the photoinitiator and inhibitor molecules are consumed. This behavior is in stark contrast to approaches such as STED or RAPID, in which the initiators are regenerated by the depletion beam. This issue can be cumbersome when fabricating closed-packed features, as the resolution declines significantly as the number of writing passes over the same region increases. For instance, Gu and coworkers reported minimum pitch of 52 nm when creating a pair of lines, however, the pitch increased to 90 nm for creating 10 lines [35]. This issue can be addressed partially by increasing the concentration of the photoinhibitor molecules in the photoresist so as there will be always enough quenching in every exposure cycle over the same volume of the material. However, to keep the size of the closely-packed features constant the exposure parameters must be adjusted in each region of the pattern.

Increasing the concentration of the photoinhibitor in a photoresist may be challenging. A typical photoinhibitor breaks into two radicals that are weakly reactive. These radicals react more efficiently with existing radicals than with monomers. With the appropriate formulation of photoinhibitors in a photoresist, the overall effect is the quenching of polymerization. However, at high enough concentrations these radicals can still drive crosslinking reactions, which puts a limit on the maximum deactivation intensity that can be used.

3.4 Photoisomerization

Absorption modulation lithography (AML) was introduced in 2009 by Menon and coworkers [25]. In AML, a thin layer of photochromic molecules is placed on top of a thin layer of photoresist. The photochromic molecules have two isomeric forms that can switch via absorption of visible or UV light (Fig. 5(a)). When one of these photochromic molecules is exposed to red light it generates an “open” ring form that absorbs strongly in the UV and is transparent in the red (Fig. 5(b)). Irradiation with UV light generates the “closed” ring form that absorbs red light strongly, but absorbs UV light with considerably less efficiency than the “open” form. To create the nanoscale structures, the photochromic layer is exposed to an interference pattern that overlaps the peaks of UV light with the nodes of red light (Fig. 5(c)). In the regions where there is red light, the layer absorbs the UV radiation, protecting the photoresist. However, in the regions with weak or no red intensity, which corresponds to the areas of the UV peaks, the UV radiation penetrates the photochromic layer to expose the underlying photoresist (Fig. 5(d))). The exposed area has subwavelength dimensions that can be decreased by increasing the ratio of the intensity of the red light over the UV. Ultimately, the diffraction of the red light determines how closely individual features can be packed in a single exposure. The use of multiple exposure cycles with different patterns allows for packing of features beyond the diffraction limit. A major advantage of AML is that the excitation is decoupled from the inhibition. Although such decoupling is found in other techniques, in AML the photoinitiators are in a different region than the inhibitors, which are also regenerated.

 

Fig. 5. (a) Structures of the open- and closed-ring isomers of 1,2-bis(5,5′-dimethyl-2,2’-bithiophen-yl)perfluorocyclopent-1-ene. (b) Absorbance spectra of the compound in the open and closed forms in hexane. (c) Deep subwavelength patterning using absorbance modulation. The photochromic layer is illuminated by two overlapping standing waves with periods of 350 nm (λ₂ = 633 nm) and 170 nm (λ₁ = 325 nm), respectively. Simulations of the transmitted light at λ₁ supported narrow lines in which the peaks of the λ₁ standing wave coincide with the nodes of the λ₂ standing wave. (d) Scanning electron micrograph of lines exposed using AML. Although the photoresist is underexposed, the lines represent a recording of an aerial image that is consistent with simulation. Reproduced with permission from ref. [25], copyright American Association for the Advancement of Science.

Download Full Size | PDF

AML also has some inherent disadvantages. Photochromic molecules typically have a long switching time. This time is further increased if the molecules are in a solid matrix. Also, these molecules can only be cycled between the two isomers a finite number of times. The last issue needs to be considered in the case of dense patterning, in which the photochromic molecules must be cycled many times. So far, AML has only been demonstrated for 2D fabrication. The difficulty in implementing the technique with 3D fabrication relies on the fact that the photochromic overlayer needs to be in close contact with the resist.

In another approach, Wegener and coworkers demonstrated surface 2C lithography based on cis-trans photoisomerization of intermediate photoenol species derived from α-methyl benzaldehydes. This approach was used to demonstrate linewidths as small as 60 nm and line gratings with a lateral resolution of 100 nm [37]. Recently, the same group also used photochromic molecules as crosslinkers for 2CL printing using two-photon excitation [38]. Their photoresist was based on methyl methacrylate copolymers bearing spirothiopyran (SP) side groups. Upon 820 nm laser excitation the side groups can transition from the thermodynamically stable SP to its open merocyanine (MC) form. The MC form can interact with other SP or MC moieties from the same or neighboring chains and create supramolecular links. Under 620 nm CW laser irradiation the MC form converts to the SP form, inhibiting crosslinking. The addition of a phase mask in the depletion beam resulted in a 2-fold improvement in linewidth, down to 31 nm. Despite the small linewidth attained, the two-color photochromic approach showed limited subdiffraction resolution capability. It is believed that swelling induced during the development causes microbridging between adjacent lines, as well as detachment from the surface.

3.5 Triplet-state absorption

Most of the photoinitiators that have been successfully used in MAP form radicals some time after undergoing ISC to a triplet state. In the case in which the triplet is long-lived, another 2CL approach is to deplete the triplet state before irreversible chemistry occurs. This condition appears to be met in the photoinitiator isopropyl thioxanthone (ITX) [39]. Nonlinear excitation at 810 nm and depletion with a phase-shaped CW beam at 532 nm resulted in polymerized lines with linewidths as small as 65 nm, i.e. 1/12.5 of the excitation beam wavelength. It should be noted that in this first demonstration of 2CL using ITX-based photoresists, it was believed that the depletion mechanism was STED [39]. However, the fluorescence of ITX is weak at 532 nm. Later ultrafast pump-probe experiments indicated that ESA dominates over stimulated emission at 532 nm, and for longer wavelengths as well [40].

Harke et al. performed time-resolved experiments on ITX-containing photoresists using either linear or two-photon excitation for initiation, and a 642 nm CW laser for inhibition [41]. When using linear excitation with a 405 nm CW laser, it was found that the timescale of the inhibition was much longer than the fluorescence lifetime. Thus, STED should have negligible contribution on polymerization inhibition for ITX. Instead, the timescale of inhibition was to found to be within the same range as a typical triplet lifetime (ca. 1-10 µs). The hypothesized inhibition mechanism involves excitation from the reactive T₁ to a higher triplet state T_N. Interestingly, two-photon excitation at 760 nm resulted in significant shortening of the inhibition time window (< 0.2 µs), suggesting that the absorption of the 760 nm light also causes a T₁ - T_N transition, adding an extra channel to depopulate T₁ more quickly. The nature of the depopulation channel still needs to be clarified. Internal conversion in the triplet manifold (T_N → T₁) should be a fast process (∼ hundreds of femtoseconds). Therefore, an efficient transition from the triplet manifold is required to observe polymerization inhibition under steady-state conditions. A possible mechanism, for example, could rely on reverse ISC [42].

The approach of inhibiting polymerization through triplet-state absorption has the benefit of the magnitude of the triplet lifetime with respect to the singlet-state lifetime. STED requires high intensities of the depletion beam and pulsed excitation that should cover as much as possible of the fluorescence lifetime of the dye. The triplet lifetime though is typically two or more orders longer than the fluorescence lifetime. Similar to photo-induced electron back- transfer, the deactivation intensity needed to achieve the same inhibition rate will be low, so that there is no need for pulsed excitation. The drawbacks of this approach are the limited fabrication speed for point-to-point fabrication due to the slow deactivation mechanism, as well as the fact that the initiation of polymerization is not decoupled from the inhibition process.

4. Future prospects for 2CL in MAP

2CL approaches are well established as a means of improving the quality of individually fabricated 3D structures in MAP and hold great promise for improving the resolution. Improving the resolution will be a decisive step toward the transition of 2CL from being limited to research use to being a powerful industrial tool. The Achilles heel of all 2CL approaches is that there is a competition between polymerization and deactivation. In STED microscopy, Eq. (3) can lead to resolution improvement by simply employing enough deactivation intensity to overcome the saturation intensity as the majority of excited molecules must be deactivated to cause a substantial improvement in resolution. In 2CL, it is only necessary to deactivate enough molecules to drive the system back below the exposure threshold. However, if the deactivated efficiency is not 100%, the remaining undeactivated photoinitiator molecules will eventually form radicals and some crosslinking will occur. This background crosslinking limits resolution when trying to fabricate structures on as tight a pitch as possible. Because the minimum distance between features that can be patterned simultaneously is determined in part by the deactivation wavelength, multiple patterning steps are required to created tightly-packed features.

Exposing the same area repeatedly causes the background crosslinking to accumulate, and concomitantly decreases the attainable resolution, even when the degree of excitation in undesired regions is quite small. One possible approach for circumventing this problem is three-color lithography (3CL). In 3CL, a first color of light excites the photoinitiator molecules to an unreactive state that can be deactivated by a second color of light. Ultimately, a third color of light can take the remaining excited molecules to a reactive state that leads to crosslinking, decoupling deactivation and crosslinking [32,43]. This approach is expected to result in significantly higher resolution than 2CL [32].

The mechanisms for initiation and inhibition for many of the materials used in 2CL remain poorly understood. For example, it is not clear how dye molecules can initiate radical polymerization [33]. Also, there is a lack of knowledge of the order of nonlinearity of 2CL photoinitiators, and of how many photons must be absorbed before the system reaches the reactive state. If photosensitizers are to be improved further, it is essential to understand all of the photophysical and photochemical mechanisms for sensitization and deactivation of different classes of materials [44]. In this direction, a number of approaches have focused on determining the effective order of the nonlinear absorption of multiphoton photoresists [45–48]. Still, there is an imperative need to introduce new methodologies for performing additional comprehensive, photophysical studies on sensitizer systems for 2CL.

Another challenge for 2CL is increasing the fabrication speed. The fabrication speed is inversely proportional to the exposure dose, so addressing this issue therefore requires developing improved photoinitiators with higher MPA cross-sections and high reactivity. Furthermore, to attain the highest possible resolution it is necessary to push the deactivation color to the shortest possible wavelength, so that the dark region in the center of the deactivating beam will have the smallest possible dimensions.

Finally, an important concern for MAP-based 2CL is the mechanical rigidity that the structures need to have to prevent disintegration during the development procedure. Addressing this issue may require techniques such as supercritical drying or the extension of the two-color approach to other materials, such as hybrid organic/inorganic composites.

5. Conclusion

Two-color approaches in photolithography have enabled substantial resolution enhancement in MAP and hold great promise for improving the resolution down to the 20 nm level. Despite the similarity of multicolor lithography to STED microscopy, understanding and controlling of the physical and chemical properties of photoresists poses new challenges. Addressing these challenges successfully will require the development of new approaches to improving both two-color materials and techniques. With further advances, multicolor techniques may become the leading tabletop platform for performing 3D lithography with resolution that rivals that of more expensive solutions, such as electron-beam lithography.