Chemical-etch-assisted growth of epitaxial zinc oxide
E. J. Adles and D. E. Aspnes
Department of Physics, NC State University, Raleigh, NC 27695-8202∗
(Dated: October 12, 2009)
We use real-time spectroscopic polarimetric observations of growth and a chemical model derived
therefrom, to develop a method of growing dense, two-dimensional zinc oxide epitaxially on sapphire
by metalorganic chemical vapor deposition. Particulate zinc oxide formed in the gas phase is used
to advantage as the deposition source. Our real-time data provide unequivocal evidence that: a
seed layer is required; unwanted fractions of ZnO are deposited; but these fractions can be removed
by cycling between brief periods of net deposition and etching. The transition between deposition
and etching occurs with zinc precursor concentrations that only differ by 13%. These processes are
understood by considering the chemistry involved.
arXiv:0910.2199v1 [cond-mat.mtrl-sci] 12 Oct 2009
Zinc oxide (ZnO) is a transparent conducting oxide with a room-temperature band gap
of 3.37 eV. It is currently under intense investigation for magneto-optic applications and as
a cheap replacement for optical and optoelectronic devices currently depending on gallium
and indium. [1, 2] Bulk, epitaxial, and nanostructured ZnO have been grown by a vari-
ety of methods including metalorganic vapor phase epitaxy, metalorganic chemical vapor
deposition (MOCVD), molecular beam epitaxy, pulsed laser deposition, and vapor-liquid-
solid processes. [1, 2, 3] Of these processes MOCVD has several advantages, including the
production of high-quality films through fine-tuning of various processing parameters, sim-
pler and less costly equipment, scalability, and higher throughput compared to conventional
physical-vapor-deposition techniques. 
Here we address two issues that cause difficulties in the growth of zinc oxide by MOCVD.
The common zinc precursors diethylzinc (DEZ) and dimethylzinc readily react in the gas
phase with oxidizing species such as O2, NO, and N2O to create particles of sizes of the
order of hundreds of nanometers or larger (the so-called particle problem).  Adverse
effects of particulate ZnO in the gas phase range from deposition of poor-quality material to
the clogging of process lines leading to reactor downtime. Second, deposition of ZnO often
yields granular or columnar structures and surface roughness of a few to a few tens of nm.
 Some of the recent attempts to address these issues include the use of exotic precursors
such as Zn(TTA)2TMEDA and Zn(TMHD)2;[4, 6] alcohols as oxidizing agents; modified
reactor designs that separate precursors and effectively transform the growth process into
alternating layer epitaxy; atmospheric pressure deposition;[8, 9] and variation of the VI/II
In this work we report results obtained by real-time spectroscopic polarimetry that pro-
vide insights into the chemistry and growth of ZnO by MOCVD. We take advantage of
these insights to develop an efficient process where the consequences of the above difficul-
ties are minimized. Minimum modification of our MOCVD reactor is required. We show
that particulate ZnO in the gas phase, when properly managed, provides a useful source for
Our deposition system is a modified Emcore GS-3300 MOCVD reactor with an inte-
grated real-time spectroscopic polarimeter for in-situ analysis and control of growth. The
polarimeter measures the relative reflectance and the complex reflectance ratio from 240 to
840 nm at a 4 Hz rate, allowing growth to be followed on this time scale. Details of the
growth system and optical diagnostics are discussed in previous work. [11, 12] The only
modification specific to ZnO growth was the addition of a down-tube to keep the precursors
separated as long as possible. The down-tube terminated approximately 2.5 cm above the
sample, a spacing determined by the need to maintain an unobstructed light path for the
A qualitative understanding of the connection between ZnO density and the <?> spectra
measured by polarimetry can be achieved by means of the linear expansion of the Fresnel
equations for small d/λ, expressed in pseudo-dielectric function form :
<?> =<?1> +i <?2>
?s(?s− ?o)(?o− ?a)
where ?s, ?o, and ?a = 1 are the dielectric functions of the sapphire substrate, the ZnO
overlayer, and the ambient, respectively, and φ = 70◦is the angle of incidence. In the
region of transparency of ZnO we have ?s≈ 3.12 and ?o≈ 4.00, neglecting birefringence
and dispersion. Thus if bulk ZnO is being deposited, ?o> ?sand Eq. (1) shows that in the
transparent region <?2> will be negative. Conversely, if <?2> is positive, then ?o< ?s, i.e.,
the deposited ZnO contains a significant fraction of voids.
We report the results of two experiments where ZnO was deposited on (0001) sapphire
(Al2O3) substrates. In run #1, we illustrate various aspects of growth using DEZ and O2as
the zinc precursor and oxidizer, respectively. Ultrahigh purity nitrogen (UHP N2) was used
as the carrier gas for DEZ. The flow rate of the carrier gas was manually adjusted between
10 to 100 sccm in response to real-time values of <?2>. All other process settings remained
constant. N2boiled off from liquid N2was used as a pump ballast, to maintain chamber
pressure, and to prevent deposition on the optical viewports. The growth parameters were:
substrate temperature 391◦C; chamber pressure 80 Torr; O2flow rate 10 sccm; main N2flow
rate 1.5 slm; DEZ bubbler temperature -10◦C, and UHP N2(DEZ carrier gas) 10-100 sccm.
No pressure controller was used in the DEZ lines, so the pressure in the DEZ bubbler was
established by the chamber pressure.
In run #2 we apply the information obtained to achieve two-dimensional growth of dense
material. In addition to DEZ and O2, 3% NO in N2was added to attempt p-type doping
by nitrogen substitution. Parameters here were: substrate temperature 490◦C; chamber
pressure 140 Torr; O2flow rate 20 sccm; 3% NO in N2flow rate 100 sccm; main N2flow rate
FIG. 1: View inside the growth chamber showing a typical vertical distribution of ZnO particles.
The heater, substrate, and down-tube are visible at the bottom, center and top, respectively. The
polarimeter beam entering at the upper right is made visible by scattering from the particles,
indicating sizes of the order of hundreds of nm. (color online)
1.5 slm; DEZ bubbler temperature 15◦C; and UHP N2(DEZ carrier gas) flow rate 50 sccm.
In contrast to run #1, we used a pressure controller in the DEZ line after the bubbler to set
the pressure in the bubbler. This allowed us to cycle between short periods of deposition
and removal by varying the pressure in the bubbler. This proved to be the key that allowed
us to grow dense material. Specifically, we cycled the pressure between 450 and 400 Torr,
corresponding to DEZ flow rates of 1 and 1.125 sccm, respectively. The cycle period was
approximately 30 seconds, limited by the response time of the pressure controller.
Figure 1 shows a typical vertical distribution of ZnO particles above the sample surface
that results from the reaction of DEZ and O2in the gas phase. Particles are made visible by
light scattered from the incoming polarimeter beam. The scattering indicates ZnO particle
sizes of the order of hundreds of nm. Of particular interest is the sharp cutoff of scattered
light approximately 5 mm above the sample. Within this 5 mm the particle size has obviously
dropped below the scattering threshold, indicating sublimation or etching. Since the same
process is likely to be occurring at the substrate, we conclude that fairly large exchange
currents must be present, although with the net result that ZnO is deposited on the substrate.
Thus two exchange currents must be managed: that originating with the particles, and the
second from the material deposited on the substrate. In our system the susceptor provides
the necessary heat to drive the reactions. More generally a separate heating element could
be used for independent control.
With the basic conditions established, we now consider ZnO deposition in detail. Figures
2 show <?1> and <?2> as a function of energy in eV, at four different times into run
#1. These spectra correspond approximately to reflection and absorption, respectively.
In Fig. 2(a) <?> describes the dielectric response of the bare Al2O3 substrate, which is
transparent and highly polished. Therefore, no absorption is seen and <?1> is nearly flat
across the visible spectrum. The presence of any ZnO would result in structure near 3.3 eV
as seen in Figs. 2(b)-(d). From the lower pane of Fig. 2(a) we see that carrier gas flow rates
of 10 and 50 sccm yield no initial deposition. At 100 sccm deposition begins, as indicated
by the structure developed near 3.3 eV in Fig. 2(b). Here, <?2> is negative below 3.3 eV.
From Eq. (1) we conclude that the deposited material has a refractive index n greater than
that of Al2O3, which is characteristic of bulk ZnO. Figure 2(c) shows that reducing the flow
rate to 10 sccm results in removal of the deposited ZnO. This is evident by the reduction
in structure near 3.3 eV, and that <?2> below 3 eV has become positive. This indicates
that n of the overlayer is now less than that of Al2O3, ie., that the ZnO has become porous.
We interpret this as preferential sublimation or etching of either defective material or of
crystal orientations other than that epitaxially grown on the substrate. Figure 2(d) shows
that returning to 50 sccm results in deposition, whereas no deposition initially occurred at
50 sccm as seen in Fig. 2(a). In addition, the <?2> spectrum shows that this new material
is filling in the voids. From this we deduce that a seed or striking layer must be established
for subsequent growth to occur, and that a cyclic pattern of removal and deposition may
result in improved material.
Figures 3(a)-(d) show the progression of <?> during run #2, where the cyclic growth
strategy suggested by the above results was implemented. In Figs. 3(a)-(c) the structure near
3.3 eV is increasingly better defined and <?2> increasingly negative after successive cycles.
In Fig. 3(d), obtained after run #2 ended and the sample cooled to room temperature, we see
a well-defined peak in <?2> near 3.3 eV. AFM measurements of the resulting ZnO, shown
in Fig. 4(a), reveal a RMS roughness of 3.93 nm over a 2x2 µm2area. For comparison, Fig.
4(b) shows AFM measurements of a sample prepared without cycling the DEZ. The RMS
roughness here is 12.36 nm over a 2x2 µm2area. Ex-situ ellipsometric measurements given
in Fig. 5 show that the room-temperature <?> spectra can be modeled as a bulk ZnO layer
54 nm thick. Thus a dense layer of ZnO was achieved. A movie of the realtime spectra from
the two runs is available online at http://www4.ncsu.edu/ ejadles.