Metal-organic gas phase epitaxy

The metal-organic vapor phase epitaxy (engl. Metal organic chemical vapor phase epitaxy , MOVPE , and organo-metallic vapor phase epitaxy , OMVPE ) is an epitaxial process for producing crystalline layers. It is related to the used equipment identical to the metal organic chemical vapor deposition (Engl. Metal organic chemical vapor deposition , MOCVD ), the terms MOVPE, MOCVD and OMVPE in connection semiconductor region are generally used for the same processes. The MOCVD describes each deposition with the process, the MOVPE and OMVPE only the epitaxy, i.e. the (single) crystalline growth on a crystalline base. In contrast to molecular beam epitaxy (MBE), the crystals do not grow in a high vacuum , but in a low vacuum (20 to 1000 hPa ).

MOVPE is the most important manufacturing process for III-V compound semiconductors , especially for gallium nitride (GaN) based semiconductors , which is now the most important base material for blue, white and green LEDs .

Organometallic raw materials

Since the starting materials for compound semiconductors are often metals, they cannot be introduced into the gas phase in elemental form at low temperatures. Therefore, with this epitaxy method, the starting materials are made available in the form of organometallic compounds (e.g. trimethylgallium ) and hydrides (e.g. ammonia , phosphine , arsine ). The advantage of these compounds is a moderate vapor pressure at room temperature, so that they can be vaporized and transported through pipelines close to standard conditions .

The organometallic compounds are stored in so-called bubblers ( gas washing bottles based on the construction principle ) and form a saturated vapor above the liquid or solid, which is transported into the reaction chamber ( reactor ) with a flowing carrier gas (usually hydrogen or nitrogen, formerly also argon ) . The bubblers are located in thermostats, with the help of which they are kept at a constant temperature in order to obtain a defined constant vapor pressure of the organometallic substance . The molar flow of the organometallic can be determined from the total pressure and the flow of the carrier gas : ${\ displaystyle p _ {\ mathrm {met}}}$ ${\ displaystyle p _ {\ mathrm {dead}}}$ ${\ displaystyle f _ {\ mathrm {t}}}$ ${\ displaystyle f _ {\ mathrm {met}}}$

{\ displaystyle f _ {\ mathrm {met}} = f _ {\ mathrm {t}} \ cdot {\ frac {p _ {\ mathrm {met}}} {p _ {\ mathrm {tot}}}}}

Reaction processes

The gross reaction formula of trimethylgallium (CH ₃ ) ₃ Ga and ammonia NH ₃ during the growth of gallium nitride can be as

{\ displaystyle \ left (\ mathrm {CH} _ {3} \ right) _ {3} \ mathrm {Ga} + \ mathrm {NH} _ {3} \ rightarrow \ mathrm {GaN _ {\, (fixed)} +3 \ \ mathrm {CH} _ {4 \, (gas)}}}

to be written. However, this reaction is a great simplification of the actual situation before and during the crystal growth. For example, preliminary reactions take place between the starting materials in the gas phase and inert adducts are often formed . The possible individual reactions are varied depending on the type of starting materials and carrier gases used and can only be vaguely predicted because of the difficult-to-determine catalytic properties of the various surfaces of the MOCVD reactor.

Due to the principle-related presence of large amounts of carbon and hydrogen, small amounts of these substances are always built into the semiconductor crystal. Hydrogen often passivates the acceptors required for the p-type line , but can usually be removed simply by tempering in an inert gas atmosphere or in a vacuum. Carbon is usually not a problem and is used specifically for p-type doping during gallium arsenide growth.

Due to the high purity requirements of semiconductors (typ. <10 ppb ), all starting materials must be in a highly pure form. The carrier gases hydrogen ( palladium cell ), nitrogen and argon ( getter filter ) can easily be represented in a highly pure form (9N = 99.9999999%). Also for the hydrides nowadays there are efficient getter the common contaminants (eg. B. , , ) almost completely removed. The organometallic agents are still one of the main sources of contamination, in particular oxygen , despite complex manufacturing processes . ${\ displaystyle \ mathrm {H_ {2} O}}$ ${\ displaystyle \ mathrm {O_ {2}}}$ ${\ displaystyle \ mathrm {CH_ {4}}}$

Growth process

Subdivision of the growth areas in the MOVPE

For layer growth , the reactants diffuse from the gas flow to the substrate surface , where they are incorporated into the crystal. At low temperatures, the incorporation of the reactants is determined by their decomposition. This is the kinetically controlled area. Since the decomposition of the starting materials or surface reactions have an exponential dependence on the temperature, the growth rate in this area is very temperature-dependent and therefore difficult to control. At higher temperatures, the growth is again limited by the replenishment, i.e. the rate of diffusion . However, as a first approximation, the diffusion is not temperature-dependent. Therefore, work is usually carried out in the diffusion- controlled area. At higher temperatures, growth-inhibiting pre-reactions occur to a greater extent or the vapor pressure of the semiconductor becomes so high that the growth rate is reduced again ( desorption ). This reduction in the growth rate also has an exponential dependence on temperature. Therefore, this area is also difficult to control and is avoided.

The surface processes during growth play another crucial role. The processes can be divided into the transport of the reactants to the surface, chemical reactions and adsorption on the surface, surface kinetic processes and desorption and the removal of the reactants. As already mentioned above, the aim is for growth to take place in a diffusion-limited manner, i.e. transport-limited. The growth is then only limited by the diffusion of the starting materials to the substrate or by the removal of the products from the substrate. In the kinetically limited area, for. B. the desorption of the reaction products may be hindered. Then, due to the incomplete removal of the remaining reactants, increased carbon incorporation is to be expected. For the normally desired smooth surfaces, sufficient mobility of the starting materials on the surface is also important in order to achieve step growth .

Surface processes during layer growth at MOVPE

In addition to the temperature and the total pressure in the reactor, the partial pressure of the reactants used and their partial pressure ratios are essential for the growth process. Among other things, this is decisive for the stoichiometry and the growth mode, i.e. i.e., whether island growth or step growth occurs. These parameters can be used to influence the growth rates of various crystal facets but also the incorporation of impurities. If, in addition, a strained ternary layer is grown, it can, depending on the material combination and growth parameters , in the Frank-van-der-Merwe mode as a two-dimensional layer, in the Stranski-Krastanow growth mode as a wetting layer with subsequent three-dimensional island growth or Grow up directly as three-dimensional islands in the Volmer-Weber growth fashion. Using the Stranski-Krastanow mode, self-organized quantum dots are nowadays often used , preferably in the In (Ga) As / GaAs system for applications such as e.g. B. Quantum dot laser grown.

Advantages and disadvantages of the MOVPE

With the MOVPE, the semiconductor crystal layers that are important for the function of the components can be reproducibly grown to a monolayer (<2.5 Å ). Typical growth rates are between 0.1 nm / s and 1 nm / s and thus higher than with MBE. The method was mainly used in the 1980s by the simple possibility of using phosphorus-based semiconductor crystals such as B. promoted to grow indium phosphide . Until then, this was not possible or only possible to a limited extent with the MBE. In the 1990s, the blue LED based on gallium nitride and, to a lesser extent, the growing market for GaAs and InP- based components for low- dispersion and low-loss data communication around 1310 and 1550 nm via fiber optic cables and microwave applications for mobile phones and Military applications ( radar ) triggered a boom for MOVPE technology. GaN in particular cannot be produced with the MBE in sufficient quality and quantity for LEDs. Due to the simple scalability of the systems and processes (from simple 2-inch single wafer systems to 95 × 2-inch or 25 × 4-inch wafers), it is ideally suited for mass production. By dispensing with the high vacuum equipment required for MBE, MOVPE technology is relatively inexpensive and easy to maintain.

The main cost factors are the expensive, high-purity raw materials and the low material efficiency compared to MBE. By working with elemental compounds, in contrast to the MBE, relatively large amounts of foreign atoms (C, O, H) are always incorporated into the crystal and therefore no semiconductor crystals as pure as with the MBE can be produced.

literature

Deodatta V. Shenai-Khatkhate, Randall J. Goyette, Ronald L. DiCarlo Jr., Gregory Dripps: Environment, health and safety issues for sources used in MOVPE growth of compound semiconductors . In: Journal of Crystal Growth . tape 272 , no. 1–4 , 2004, pp. 816-821 , doi : 10.1016 / j.jcrysgro.2004.09.007 .