Paul F. Lyman, PhD , University of Wisconsin (Milwaukee)
To perform ZnO ALD we have designed and constructed a viscous-flow ALD reactor. The design is based heavily upon that of Elam, Groner, and George. The ALD reactor was constructed of a stainless steel tube of 4.5 inch outside diameter and 24 inch length. A commercial proportional, integral, differential (PID) temperature controller (CAL 9300) was used to control the power input to an external heating jacket to heat the reactor tube. A chromel-alumel thermocouple was attached just below the substrate holder to monitor the ALD growth temperature. Moreover this ALD reactor was attached to an existing UHV analysis chamber. Ultrahigh pure nitrogen gas (99.99%, Praxair Inc.) was used as carrier gas. Before entering the ALD reactor, the carrier gas was purified using an Entergris GateKeeper gas purifier. Nitrogen gas flow rate of 120 sccm was controlled using a MKS Instrument Inc. mass flow controller. ZnO ALD was performed at a reactor pressure of 1 torr. The pressure was maintained using a MKS Instrument Inc. 253A throttle valve, MKS Instrument Inc. 265C pressure controller, and a dual-stage rotary vane mechanical pump (Alcatel 2015-C2). The process pressure was monitored by a Baratron capacitor manometer (MKS Instrument Inc.).
Diethyl zinc (DEZ) (96%) from Strem Chemical Inc. and Fisher Scientific deionized water were exposed alternately to achieve ZnO ALD. The DEZ and water were both kept at room temperature and dosed directly from the precursor delivery cylinder. All the precursor lines of the viscous-flow ALD reactor were equipped with pneumatically actuated diaphragm valves. These diaphragm valves were controlled by a personal computer. LabView was used to control the pulse timing of these valves. ZnO ALD was successfully performed by alternating dose of DEZ and water with a pulse sequence t1-t2-t3-t4, where t1 is DEZ pulse time, t2 is nitrogen purge time, t3 is water pulse time, and t4 is nitrogen purge time. Needle valves attached with the precursor lines were used to adjust the transient pressure amplitude to a value of ~ 0.1 torr during the precursor pulse.
In-situ QCM was employed to investigate the self-limiting behavior of ZnO ALD. As a function of DEZ pulse length (at a constant temperature of 130˚C), the growth rate of ZnO film initially increases rapidly but then becomes constant, for pulse times above 2 sec. Assuming a density of ZnO of 5.6 gm/cm3, each cycle corresponds to ~2.3 Å. The behavior as a function of water pulse time is similar. Next we investigated the dependence of the average growth rate per cycle of ZnO ALD on reaction temperature, and found that there is a temperature “window” in which ALD growth shows near-ideal behavior. At lower temperatures, insufficient thermal energy is available for the reactions to proceed. At higher temperatures adsorbed molecules and/or the previously deposited film constituents may decompose. In the present case, we found a window from about 130˚ to 170˚C where good, self-limited ZnO growth may take place. Using in-situ QCM we have observed a linear step-like ZnO ALD growth, which is consistent with earlier work in the literature.
Ex-situ AFM study was performed on the ZnO ALD films deposited on both oxidized Si(100) and MgO(111) polar substrates. All experiments were performed on samples grown at 130˚C and ALD cycle time 3-5-2-5 s. AFM scan area 1x1 _m was used to find the RMS surface roughness value of the ZnO ALD films. After 200 cycles RMS roughness value for films on a Si(100) substrate was found to be ~ 1.5 nm, whereas for the MgO(111) substrate it was ~2.4 nm. The RMS surface roughness of ZnO thin films on the MgO(111) substrate increases to ~ 4.4 nm for 600 ALD cycles and ~ 5.3 nm for 1000 cycles.
In all, these efforts have established that we can grow high-quality ZnO films using ALD under a range of processing conditions. Further work is underway to improve the crystallinity of these films and to explore (Zn,Mg)O alloys.