Direct plasma-enhanced atomic layer deposition of aluminum nitride for water permeation barriers Open Access

Aluminum nitride (AlN) has been investigated for its use, e.g., in light emitting diodes,^1,2 high power,^3,4 and microelectromechanical devices^5–7 due to its wide bandgap, high dielectric constant, and piezoelectric properties. To achieve high quality, AlN has typically been grown by chemical vapor deposition (CVD)^3,4 at high temperatures, which, however, limits the use of materials requiring low process temperatures. To reduce the growth temperature, atomic layer deposition (ALD) is a versatile technique, where depositions are possible for temperatures >350 °C (Refs. 8 and 9) or >200 °C (Refs. 5 and 10–13) for plasma enhancement (PE).

AlN depositions may also be interesting for oxygen-sensitive materials, like organic semiconductors. Organic electronic devices (OEDs) enable low-cost, high-volume production of transistors, light-emitting diodes, and solar cells through vacuum or solution processing. The state-of-the-art printing technology and the absence of high temperature processes enable a wide variety of substrate materials, such as polymer films or metal foils. Furthermore, most materials used in organic electronics are intrinsically flexible, do not use rare materials, and have tunable properties. However, unprotected OEDs are degrading in ambient conditions due to layer or interface degeneration.^14–16 Therefore, a high-quality moisture barrier is needed.

ALD can produce conformal¹⁷ and uniform films with low defect densities. Water vapor transmission rates in the range of 10⁻² mg/m²/day were reported for Al₂O₃ at 120 °C deposition temperature on bare polyethylene naphthalate (PEN) films.¹⁸ 

A temperature below 80 °C is recommended for direct deposition on organic light emitting devices to prevent degradation as with decreasing temperature, the layer density decreases and the impurity levels increase.¹⁹ To improve the layer quality, additional energy by plasma,^20–23 UV light,^24,25 or flash lamp annealing²⁶ can be applied.

While plasma enhancement can improve the film density, adhesion, and composition, oxygen gas plasmas, as used for the common barrier material Al₂O₃, are harmful for organic devices. The activated species (ions, radicals, and metastable molecules) react with the surface and may combust the organic layer. The etching effects also hinder the film growth, until a plasma protection layer is formed.²⁷ Upon replacing oxygen with nitrogen and hydrogen for nitride deposition, evidently no oxidation effects are expected due to the reducing properties of the excited species.

For the use of nitride as a permeation barrier, only few materials can deliver an appropriate optical transparency. The most commonly used one is silicon nitride, which is deposited by CVD.^28–30 However, layer thicknesses in the range of several hundred nanometers are required to lower the defect density and to achieve sufficient barrier performance. To enhance the quality and lower the thickness, Andringa et al. deposited plasma enhanced ALD (PEALD)-SiN_x on PET films with suitable low intrinsic water permeation rates in the range of 10⁻⁶ g/m²/day at 25 °C/50% RH for thicknesses between 10 and 40 nm.³¹ An alternative to consider is AlN, which has similar properties (optical transparent, high bandgap, and electrical insulating) like SiN and the used precursor trimethylaluminum (TMA) is well known and more efficient than bis(tertiary-butylamino)silane (BTBAS).

In previous studies, we achieved low water permeation rates with oxide processes.^32,33 For further improvement of the water barrier and to decrease the possible damages to organic devices,³⁴ we developed an aluminum nitride process. Films were deposited at 80 °C and 200 °C and their growth and properties such as refractive index and composition and electrical characteristics are presented and discussed. The resulting film properties obtained by direct plasma deposition are compared with thermal ALD and remote PEALD references. Finally, the suitability of AlN as a water permeation barrier is shown.

Silicon with native oxide, platinum-coated silicon, and planarized PEN films with protecting films (125 μm, Optfine^® PQA1M from DuPond Teijin) were used as substrates. The film substrates were degassed at 70 °C in dry air for 5 h and stored in a glovebox with O₂ and H₂O levels below 1 ppm. To characterize electrical properties, the metal-insulator-metal (MIM) capacitors were fabricated by electron beam evaporation of titanium dots on top of AlN-coated Pt substrates through a shadow mask.

AlN films were grown using a TFS 500 from Beneq Oy., equipped with a capacitively coupled plasma source from FAP GmbH. A schematic of the deposition chamber can be seen in Fig. 1(a). The electrode distance is 10 mm, electrode diameter is 180 mm, and the excitation frequency is 13.56 MHz. The rf power was set to 120 W.

Al(CH₃)₃ (TMA) was used as a precursor and H₂/N₂ plasma as a second reactant. Argon was used as the electrode purge and plasma gas. N₂ (110 Pa) and Ar (50 Pa) flowed through the reactor permanently, while H₂ (70 Pa) was pulsed via a precursor inlet starting one second before the plasma power was switched on. The pressure increase of the TMA pulse was 5 Pa. Prior to every process, substrates were heated up for 30 min to either 80 °C or 200 °C. A process scheme is sketched in Fig. 1(b).

Optical properties were measured using a variable angle spectral ellipsometer (J. A. Woollam Inc.) using a wavelength range from 350 to 1000 nm. By employing a Cauchy model, film thicknesses and refractive indexes (at 500 nm) were evaluated.

Film composition was determined ex situ by an x-ray photoelectron spectroscopy (XPS) unit from Omicron Nanotechnology GmbH with Mg-Kα radiation. Prior to the measurements, the samples were sputter-cleaned with a 4 keV argon ion beam for 60 s. The layer composition was determined using the peak areas later.³⁵ 

Surface topography was measured by atomic force microscopy (NanoScope IIIm from digital instruments). Scanning electron microscopy was performed by using a Hitachi S4700.

A Keithley 4200 semiconductor characterization system was used for electrical measurements. The MIM-capacitance values were determined at 10 kHz with 30 mV amplitude and 0 V bias.

Water permeation rates were obtained by an electrical calcium test, which uses the conductance decrease of a metallic calcium thin-film upon water ingress for sensing. The thin-film is contacted by Al electrodes in a four-point-probe geometry and a mechanical buffer layer (C₆₀) between the barrier and the sensor protects against mechanical stress. Humid air is applied locally over each sensor with a saturated salt solution (Na₂SO₄ in water). More information on the setup and the procedure is given in Ref. 33. 25 × 25 mm² PEN samples were fixed on a silicon wafer using Kapton tape to prevent scratches and handling errors. After deposition, films were carefully released from the wafer and transported without breaking the inert atmosphere.

First, the required process times were investigated for AlN deposition at 80 °C substrate temperature. Initially, pulse durations of 0.5 s for TMA and 10 s for the plasma pulse were selected. Both purge times were fixed at 10 s. In each process, 100 cycles were deposited on silicon to obtain layers with a thickness of 10 nm. During the process, H₂ flowed permanently through the showerhead [line “electrode purge,” see Fig. 1(a)].

To check the precursor saturation, TMA and plasma pulse times were varied [see Fig. 2(a)]. For TMA pulse times between 0.5 and 1.0 s, constant layer thicknesses were observed, indicating the self-limiting growth behavior with a growth per cycle of 1 Å. Compared to the thermal alumina process at this tool, the surface coverage needs longer time to saturate, indicating lower reaction rates for TMA on NH_x terminated surfaces. For further experiments, the TMA pulse time was set to 0.5 s.

The plasma time was varied between 3 and 10 s, as seen in Fig. 2(a). Again, a self-limiting behavior was observed between 5 and 10 s. However, for PEALD of Al₂O₃ at 80 °C in the same tool, only 1 s plasma time was needed for saturation, suggesting lower reactivity, respectively, a slower chemical kinetics for PEALD of AlN. The plasma time was fixed at 10 s for the subsequent experiments.

Finally, the purge times were varied between 5 and 20 s. As seen in Fig. 2(b), all film thicknesses are in the range between 9 and 11 nm, revealing that even a 5-s purging was sufficient to remove unreacted precursors from the deposition chamber. Further investigations in shadowed areas revealed no significant growth, indicating a high recombination rate of the plasma-activated species. Therefore, it might be possible to reduce purge times even further below 5 s.

Although the achieved results offer the potential to reduce the cycle times and to increase the deposition speed, the working point was kept at pulse times of 0.5 s TMA pulse; 10 s purge after TMA, 10 s plasma pulse, and 10 s purge after plasma step.

After 200 cycles of AlN-deposition, the film thickness were measured by ellipsometry and shown in Fig. 3. The process at 80 °C resulted in a film thickness of 21 nm and a refractive index of 1.93 (at λ = 500 nm). The corresponding values at 200 °C are 20 nm with a refractive index of 1.98 (at λ  = 500 nm). These values indicate a slightly denser structure and higher growth rate for the process at 200 °C. However, the differences are very small. Therefore, the temperature influence appears negligible. The samples were measured immediately after deposition and a second time after one week storage in ambient air. Negligible changes in Ψ-Δ-spectra indicate a high stability against air exposure (see Fig. S1 in the supplementary material⁴⁰). Compared to literature values (see Table I), the achieved refractive indices are high, specifically considering the low deposition temperatures of 80 °C.

By applying the determined optical properties to samples with different cycle numbers, a linear relation between thickness and cycle number with an average growth rate of approximately 1 Å per cycle was obtained. The extrapolation to zero showed no significant substrate-related growth, as expected for direct plasma-enhanced processes. Literature values for the growth per cycle are in the range from 0.6 to 1.2 Å (see Table I) (Fig. 4).

The surface topology of 20-nm thick films on silicon substrates are determined by AFM. While the surface was smooth (≈0.5 nm rms roughness) at low temperatures [see Fig. 5(a)], 100-nm wide and 10-nm high features were observed on top of the 200 °C sample [Fig. 5(b)]. Beside these features, the surface was as smooth as the 80 °C sample.

To determine the origin of these features and the surface quality in more detail, wafer samples were broken and the cross section was analyzed via SEM (Fig. 6). The features identified as blisterlike delamination from the silicon surface were AlN coatings with thicknesses in the range from 20 to 24 nm.

By XPS, the layer compositions were analyzed (see Fig. S2 in the supplementary material⁴⁰). After argon sputter-cleaning, aluminum nitride with low amount of oxygen and carbon were found. The low carbon content even at 80 °C indicated effective removal of precursor residuals from the growing layer. The Al:N ratio was 1.0:0.9 for both temperatures in this work, which confirms the presence of stoichiometric AlN. The oxygen impurities might be incorporated by postoxidation effects due to the ex situ measurement setup or during layer growth by water and/or oxygen contaminations from the residual partial pressure of the vacuum chamber. Compared to Bosund et al., who also used a TFS 500 in a remote CCP setup, the films grown at 80 °C exhibit excellent air stability.¹³ 

In addition, the dielectric constant of PEALD AlN was determined by measuring the capacitance for different film thicknesses, as seen in Fig. 7. Uncertainties were determined by measuring different capacitor sizes from 90 to 450 μm diameter. The plot shows a linear behavior with negligible offset, as expected for an inert platinum surface. Permittivity values in the range of 11 were determined for both temperatures. Literature values range from 7 to 8 for MIS structures^10,37 and 10 for MIM capacitors.³⁸ 

For water permeation barrier property measurements, PEN substrates were coated with 20-nm AlN at 80 °C. In parallel, samples with 20-nm Al₂O₃ (also 80 °C deposition temperature) were prepared as well as a reference. To reduce handling damages, the PEN films were taped on silicon wafer prior degassing. The protecting foils were removed immediately before deposition to keep particle contamination as low as possible. After film deposition, calcium test structures were grown by thermal evaporation and encapsulated on the back side by a glass cavity.

After preparation, the calcium test samples were exposed to a saturated Na₂SO₄ salt solution in a temperature-controlled environment. Due to moisture ingress through the barrier, conductive calcium corrodes to insulating calcium dioxide, decreasing the current over time at a constant measurement voltage. By assuming homogeneous material properties and corrosion, the I-t-slope is proportional to the permeability q of the barrier layer.

Aluminum nitride and oxide showed roughly similar barrier properties, as can be seen in Fig. 8. By determining the current slope in the steady state regime, permeation rates around 0.5 mg/m²/day were determined at 38 °C and 90% relative humidity, which is the lower limit of the applied measurement setup. These results are very comparable to earlier experiments.^32,33,39 We assume a defect-limited permeation rate due to manual handling and particle generation inside the ALD-reactor.

In this work, a CC-PEALD AlN process was investigated at 80 °C and 200 °C. We applied direct plasma-enhanced atomic layer deposition with TMA and H₂/N₂/Ar plasma to avoid combustion in the carbon-based sublayer. For plasma pulse times larger than 5 s and TMA pulse times longer than 0.5 s, saturation behavior was found for both temperatures. High growth rates and an Al:N ratio of about 1.0:0.9 with traces of oxygen were achieved even at deposition temperatures as low as 80 °C. Due to air stability, the absence of water and oxygen during the process, and the low water permeation rates, the films are usable as moisture barriers for direct encapsulation of organic electronic devices.

This work is supported by the European Social Fund and by the Free State of Saxony of the Federal Republic of Germany under Contract No. 100242184. Furthermore, the authors thank Felix Winkler for metal deposition, Sebastian Killge for SEM, and Sandra Völkel for AFM picture preparation.