Dry Lithography of Large-Area, Thin-Film Organic Semiconductors Using Frozen CO2 Resists

To address the incompatibility of organic semiconductors with traditional photolithography, an inert, frozen CO(2) resist is demonstrated that forms an in situ shadow mask. Contact with a room-temperature micro-featured stamp is used to pattern the resist. After thin film deposition, the remaining CO(2) is sublimed to lift off unwanted material. Pixel densities of 325 pixels-per-inch are shown.

A typical process fl ow for phase-change lithography is shown in Figure 1 . The resist gas is fi rst applied to a cryogenically cooled substrate where it freezes. The desired pattern can be formed in the resulting resist fi lm via localized thermal excitation; in this work, we investigate resistive heating and stamping. As explained below, stamping is preferred, and this process of pattern transfer is described in Figure 1 c. Following physical vapor deposition of the organic semiconductor or metal, the substrate temperature is raised and the resist sublimes, lifting off unwanted materials, and leaving behind only the intended pattern of organic or metallic thin fi lm. The material that is lifted off can be caught by a shutter below the sample or, in the case of a manufacturing line, the step can be performed in a separate chamber to recover the material for reuse. A video of the frozen CO 2 mask lifting off is available in the Supporting Information.
To avoid degradation of our active materials, we employ inert CO 2 as the phase-change resist. The phase diagram of CO 2 is shown in Figure 2 a. [ 23 ] In a low pressure process like thermal evaporation, the sublimation temperature of CO 2 is reduced. For example, at our operating pressure of 10 − 6 Torr, the sublimation temperature of CO 2 is roughly 90 K. Thus, for a stable lift-off mask, the substrate must be cooled to at least 85 K, which is still within the range of relatively inexpensive cooling with liquid nitrogen.
Once the CO 2 resist is patterned by selective sublimation, it is important to control the partial pressure of CO 2 in the chamber to prevent unwanted re-condensation of CO 2 vapor on patterned regions of the substrate. It is also possible to freeze other impurity gases onto the substrate, notably H 2 O, whose phase diagram is shown in Figure 2 b. [ 24 ] In previous studies of frozen CO 2 fi lms at 10 − 7 Torr, Gerakines et al. measured a water deposition rate of 2 nm h − 1 . [ 25 ] At these rates re-deposition must be considered in our experiments, but should ultimately be of little consequence in high throughput manufacturing since the acceptable background pressures of CO 2 and H 2 O increase with reduced takt time.
The ultimate resolution of this lithographic process is determined by the resist thickness, which in turn is limited by the thickness of solid CO 2 required to withstand the thermal energy carried by the incident organic or metal fi lm. [ 26 ] To estimate the minimum resist thickness, t CO 2 , of an unpatterned fi lm, we consider the balance of the heat capacities c v , enthalpy of sublimation h s , and heats of fusion h f and vaporization h v of the resist and evaporated materials: where ρ and t are the density and thickness of a material, respectively. The subscripts fi lm and CO 2 correspond to the evaporated thin fi lm to be patterned and the resist, respectively. The minimum resist thickness obtained using Equation 1 is approximately 3 μ m and 400 nm for 100 nm of deposited silver and standard organic materials, respectively. The density of the resist depends on pressure, temperature, [ 27 ] and gas fl ow rate and, as mentioned in reference [ 28 ] a more amorphous resist avoids inhomogeneity at the length scales of the crystalline domains and is preferred for greater resolution. For the operating conditions in these experiments, the density is 1.51 ± 0.15 g cm − 3 ; see the Supporting Information for a description of the interferometric technique employed to measure density and fi lm growth rates. The inherent disadvantage of phase-change lithography complicating the patterning step is the relatively large amount of thermal energy that must be supplied to overcome the heat of sublimation and remove the resist during patterning. [ 18 ] Traditional optical lithographic exposure methods would require a great deal of power at wavelengths that are not readily available, λ = 2.7 or 4.3 μ m, [ 29 ] to achieve a suitable sublimation dose in a reasonable amount of time. Thus, although a number of selective heat sources are possible, we investigated resistive heating and a stamping technique to selectively sublime regions of the CO 2 mask. Both techniques are capable of rapidly injecting a signifi cant amount of heat into the CO 2 resist.
Resistive heating was performed by applying a voltage along an indium tin oxide wire patterned on a glass substrate. At a current density of 625 kA cm − 2 , the heat dissipated by the resistive load is suffi cient to sublime the overlying frozen CO 2 . An in vacuo photograph of this arrangement is shown in Figure 3 a.
A schematic representation of the setup used for pattern transfer by stamping is shown in Figure 3 b. The process is performed using two motorized linear stages (Standa Ltd.) to allow for motion control under vacuum: a 150 mm motorized linear stage to traverse the length of the cryogenically cooled sample and another, 30 mm in length, with actuation normal to the substrate to perform the stamping operation and to adjust the focus of a digital microscope. The microscope is used for positioning the stamp relative to the substrate and to observe the sample surface throughout the process.
In a demonstration of patterning after resistive heating, an approximately 100 μ m-wide silver line is patterned by this method; see Figure 4 a. The silver wire is observed to follow the outline of the underlying indium tin oxide (ITO) wire where it is narrowest, and the heat generation is the largest. Due to the high heat fl ux and diffusion through the substrate, however, this method yields an insuffi ciently rapid sublimation process resulting in cruder defi nition and undesirable debris. For this reason, the method was abandoned in favor of the stamping technique that heats the resist rapidly and directly, rather than through the substrate.
To demonstrate pattern transfer using a stamp, arrays of 20 μ m × 50 μ m features were patterned from thin fi lms of the common organic compound tris(8-hydroxyquinolinato) aluminum (Alq 3 ). A micrograph of one of these stamps is available in the Supporting Information. This resolution is compatible with commercial OLED display production, and exceeds that which might be used in a mobile display. The CO 2 mask and the subsequent patterned   [ 23 ] and [ 24 ]. The " × " represents the process operating point of 77 K at 10 − 6 Torr. The curves are extrapolated using the Clapeyron equation. [ 33 ]  COMMUNICATION temperature grown control devices, suggesting that cooled substrate temperatures can be employed in OLED fabrication without degradation of performance. The performance and yield of the cold OLEDs was highly variable, however, and we observed a visible grey tint in the hole transport layer due to a slight coarsening in the morphology. While morphological changes might occur during low temperature depositions, we attribute the signifi cant variation in device performance to the uncontrolled condensation of water vapor or CO 2 on our substrate surface during substrate cooling and the growth of the thin fi lms. [ 21 , 22 ] This can be rectifi ed by reducing the takt time from the ∼ 1 h process used in our laboratory, and reducing the partial pressures of water and CO 2 using cold traps. [ 28 ] In addition to addressing concerns of the organic layers' growth under cold conditions, the transistor backplane of active-matrix displays must also withstand the low temperatures of the process. To verify this, a small active-matrix OLED display was removed from a digital photo frame, pumped down to high vacuum and cooled with liquid nitrogen, and then replaced in its housing and connected to its driver. There was no noticeable difference in pixel brightness, uniformity or operation aside from the seal of the passivation glass coming loose-passivation being a manufacturing step strictly after full device fabrication.
To conclude, the use of an in situ defi nable mask has been shown to be a viable alternative to patterning thin fi lms of organic semiconductors and metals at the large scale. The is a false-color surface topography showing multiple pixels while Figure 4 f shows the two-dimensional cross section. The relative heat capacity of a stamp maintained at room temperature is more than suffi cient to rapidly remove the frozen resist. To prevent abrasion and dust formation, the surface of the stamp need not make contact with the hard substrate surface if a universal burn-off step is performed to uniformly 'etch' the residual resist. [ 19 ] In this step, all of the resist is uniformly removed a suitable depth such that none remains in the areas where the desired thermally deposited fi lm is to remain.
To examine the impact of cold substrate temperatures on OLEDs, we built and tested OLEDs on substrates cooled to 112 ± 24 K. The external quantum effi ciency (EQE) versus current density, J , and electroluminescence spectrum of these devices are shown in Figure 5 . The best devices on cold substrates yielded effi ciencies comparable to the room  in heating at a rate of 225 mJ cm − 2 s − 1 until the frozen CO 2 formed the intended pattern observed from a camera mounted in situ. We used SU-8 2150 photoresist to fabricate our stamps following prior reports. [ 30 ] When spun onto silicon wafers at 3,000 RPM, thicknesses of ∼ 115 μ m were obtained. Contact photolithography resolved pillars that tapered slightly after development in propylene glycol methyl ether acetate (PGMEA). For these experiments, the tapering is not so severe as to interfere with patterning as the resist thickness is on the order of 50 μ m. The thickness and density of frozen CO 2 fi lms is measured by double interferometry, as detailed in the Supporting Information. [ 27 , 31 , 32 ] For experiments with temperatures less than the ∼ 80 K obtainable with a liquid nitrogen reservoir, a cryogenic pump was repurposed for use as a cooling source and all components are mounted onto it via an oxygen-free high-conductivity (OFHC) copper rod. All cold parts are machined out of OFHC copper and indium foil is sandwiched between all temperature-critical interfaces. A kapton encapsulated heater placed in between the substrate and substrate holder provides adequate local heating for encouraging lift-off without adding too much heat to the bulk thermal mass of the apparatus. A silicon thermal diode was attached to the copper substrate holder to approximately monitor the temperature of the sample and a cryogenic temperature controller (Lakeshore Cryotronics) is employed to manage operating temperature. Patterning is monitored using a digital microscope mounted on the stamp actuator. The repurposed cryogenic pump's compressor and cold head are briefl y turned off during the actual stamping so that the vibrations do not interfere while the stamp and resist make contact.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. dry resist material frozen directly to the surface of the substrate alleviates many of the issues of scaling up as fi ne metal masks have proven increasingly cumbersome with area. While pixel density as high as 325 pixels-per-inch has been demonstrated, there is no reason to believe this is a fundamental limitespecially knowing that nanometer-scale patterning has been demonstrated with an electron beam paired with frozen water resist. [ 18 -20 ] Patterning organics at the nanoscale may be possible pairing phase-change resists with nanoimprinting techniques. Employing a micro-featured stamp roller pipelined with the necessary cooling apparatuses, phase-change resist patterning should allow for scaling of parallel patterning beyond what the current technologies offer.

Experimental Section
Patterning was successfully demonstrated within a vacuum in the 10 − 7 to 10 − 6 Torr range for substrate temperatures between 20 K and 100 K. Cooling is achieved using either a liquid nitrogen reservoir or a repurposed cryogenic pump depending on the desired base temperature. With the substrate suffi ciently cooled, CO 2 gas (Airgas, 99.999%) is introduced to the substrate via 1/4" copper tube attached to either a mass fl ow controller or variable leak valve depending on the desired fl ow rate; the fl ow rate also being a function of temperature and pressure; see Supporting Information.
In the resistive heating experiments, 160 nm-thick ITO was patterned by traditional contact photolithography and etched with aqua regia. Copper foil tape was used to make contact from the ITO on the substrate to a ceramic power feedthrough. A sourcemeter (Keithley 2400) was used to drive 100 mA of current through 100 μ m-wide lines resulting Figure 5 . External quantum effi ciency versus current density of OLEDs grown at T = 112 ± 24 K and room temperature (a). The normalized electroluminescence spectrum is indistinguishable from the room temperature control device (b) and device thin fi lm stack (c) are also shown.