Dynamics of Surface Reactions Relevant to Chemical Vapor Deposition

Background

One major driving force behind current research in chemical vapor deposition and its variants is the need to deposit thin films that meet the demands of ever–smaller and more delicate electronic and optoelectronic devices. In many cases, to satisfy these demands, the films must be grown at low substrate temperature and must be as contaminant–free as possible. A number of experimental approaches have been utilized, either individually or in tandem, in an effort to achieve low temperature growth. Examples include the synthesis of novel organometallic precursors that cleanly decompose at low temperatures and deposition techniques, such as Plasma–Enhanced Chemical Vapor Deposition and Laser–Assisted Chemical Vapor Deposition, that exploit non–thermal means of precursor (or precursor/substrate complex) excitation.

Within the past decade, several research groups have begun experiments to explore the possibility of using seeded supersonic molecular beam technology to activate low temperature thin film deposition. One can envision three principle mechanisms by which supersonic beams could promote low temperature film growth:

Since the translational energy of incident precursor molecules can be increased substantially through seeding techniques, activation barriers to dissociative chemisorption can be surmounted. A related possibility is collision–induced dissociation. Here, a hyperthermal beam of rare gas atoms collides with physisorbed or chemisorbed precursor molecules and induces their dissociation.

Surface diffusion of adsorbed species can be enhanced due to energetic collisions of incident species (e.g., Kr, Xe, or the precursor molecule itself) with existing surface ligands. It is also possible that excess energy available from the dissociative chemisorption event is channelled into surface diffusion.

Desorption, too, can be induced by energetic collisions of incident species (e.g., Kr, Xe, or the precursor molecule itself) with existing surface ligands. This process is referred to as collision–induced desorption (CID).

In all three cases, the incident molecular beam imparts a chemically relevant amount of energy (<140 kcal mole^–1, or < 6 eV) that is utilized to overcome an activation barrier. Note that this energy regime is sufficiently low that no surface damage occurs; it is far lower than in techniques involving ions, such as sputtering, for which the energy is often greater than 100 eV.

Nearly all published work has sought to exploit the first mechanism, translational activation of dissociative chemisorption. In the work described below, we focus on the third mechanism, which involves the collision–induced desorption of ligands.

In our initial study, we are probing the collision–induced recombinative desorption of H₂ from the Si(100) surface using hyperthermal beams of Kr and Xe.

        2. Bombard surface with seeded Kr or Xe beams (~1% / 99% He).  So far, we have used 1% Kr / 99% He beams at 2 nozzle temperatures.

                          450KÞE_Kr=39kcal/mole

        3. Measure final surface coverage of hydrogen via thermal desorption spectroscopy to quantify the effect of collision–induced desorption. The integrated peak areas are a measure of the coverage.

Apparatus

The experiments are conducted in a molecular beam–surface scattering apparatus that was designed, assembled, and tested in our laboratory at San Diego State University. The testing phase was completed recently, and we have now begun our initial experiments. The six–chamber apparatus, shown below, is equipped with a triply–differentially pumped supersonic molecular beam source coupled to an ultrahigh vacuum scattering chamber. The latter chamber houses the Si(100) crystal, an Ar⁺ sputtering gun for preparing clean surfaces, an Auger spectrometer for verifying surface cleanliness through elemental analysis, and a tungsten filament used for generating hydrogen atoms. During H atom dosing, the filament is heated to 1800 K and is located ¾² in front of the Si crystal; during molecular beam bombardment of the crystal, the filament is turned off and is translated out of the beam path.

A quadrupole mass spectrometer situated along the incident beam path in the scattering chamber is employed to measure the incident beam's translational energy distribution (via time–of –flight spectroscopy) and to record thermal desorption spectra of stable, nonreactive molecules such as H₂. In future experiments, the principle detector will be a doubly–differentially pumped quadrupole mass spectrometer whose line–of–sight to the crystal is fixed at 45o with respect to the incident molecular beam. This detector will be suitable for monitoring directly scattered species and thermally desorbed species, and for recording time–of–flight spectra of scattered species. Installation of this mass spectrometer is nearly complete; we expect it to be fully functional shortly. Both mass spectrometers have been modified for pulse counting in order to boost the signal–to–noise performance.

The seeded supersonic beam is formed by expanding a gas mixture (~1% seed gas / 99% He) of ~1 atm total stagnation pressure through a 100 micron diameter nozzle orifice. Gases of variable composition can be mixed as needed in our home–built, bakeable gas manifold, which can be evacuated to 5 x 10^–8 torr to satisfy our ultrahigh purity requirements. The beam can be interrupted with a temporal resolution of ~5 msec by an electromechanical shutter located in the first differential source stage. Also located in this chamber is a variable frequency chopper wheel mounted on a linear translation stage. With the chopper wheel removed from the beam path, the beam operates in continuous mode. With the linear translator fully extended, the chopper wheel cuts the beam into pulses of 0.25% duty cycle and variable length (minimum = 8 m sec). This mode permits time–of–flight analysis of the incident beam and, when the differentially–pumped mass spectrometer is fully operational, will permit time–of–flight analysis of scattered species.

In addition to past and present group members shown below, three former undergraduates— Steve Hile, Tom Ho, and Tom Martin— assisted in the construction of the vacuum apparatus and its accompanyng electronics.

At a surface temperature of 690 K, a small amount of hydrogen was collisionally desorbed, as indicated in the graph below. This is quite surprising since one would expect negligible coupling between the surface thermal energy and the incident energy of the beam.

The results are tentative but they suggest the viability of collision–induced desorption of H₂ from the Si(100) surface even though the conditions of the experiments have not yet been optimized (with respect to beam energy and incident angle). The enhanced hydrogen removal due to the 39 kcal/mole Kr beam is highly intriguing. We expected the absolute lower energy limit to the collision–induced desorption process to be approximately equal to the thermal desorption activation barrier of 57 kcal/mole. The reason we began with a Kr beam rather than a Xe beam (which would provide greater impact energy) was simply a matter of cost. In the absence of further experiments, any explanation is purely speculative. Two possibilities include: 1) an important role of thermal excitation in promoting collision–induced desorption (cf. the isobutyl/Al(111) results of Lohokare, Crane, Dubois, and Nuzzo in J. Chem. Phys. 108, (1998)). This idea is testable by measuring the collision–induced desorption probability at different temperatures and beam energies and observing whether the two types of energy can compensate for each other. 2) A severe collision–induced distortion of the surface lowers the barrier to desorption. This idea would be difficult to prove experimentally; theoretical studies would therefore be highly beneficial. It is thought that the transition state for H₂ thermal desorption involves a distortion of the Si lattice, and the activation energy to desorption depends on the degree and type of distortion.

Future Work

	Finish work on monohydride species
	Probe collision–induced desorption of dihydride species
	Probe collision–induced desorption of H2 from silicon alloy surfaces (e.g., SiGe)

 

Time–of–Flight Methodology

Our particular implementation of time–of–flight spectroscopy is simple but unusual in that we record spectra at two or more different flight distances (i.e., chopper–to–ionizer distances). The quadrupole mass spectrometer is mounted on a linear bellows translator and thus can be moved along the beam axis by up to six inches. The advantage of this method is that it eliminates the often time–consuming problem of calibrating the electronic delay and ion flight times (because the spectra have the same delay, which can thus be subtracted mathematically). Sample spectra of a pure argon beam expanded from a 303 K nozzle are shown below. These spectra represent a "worst–case" scenario and were chosen to illustrate the power of our method. Specifically, the mass spectrometer was slightly misaligned, which had the effect of substantially increasing the delay time to 150 m sec. Nonetheless, through a relatively straightforward nonlinear fitting analysis, we were able to extract reasonable values for the flow velocity and translational temperature, 560 m sec^–1 and 2 K.

 Two unusual features of the apparatus are illustrated below.

A unique and important feature of our UHV pumping setup is the presence of refrigerant–cooled (–50° C) diffusion pump baffles designed in our lab. Base pressures in the chambers having these baffles are 1.5–2 x 10^–10 torr. The chief advantages of our baffles, when compared to the conventional liquid nitrogen versions, are: 1) negligible operational costs because consumption of liquid nitrogen has been eliminated (except for the small amount used occasionally in the auxiliary trap mentioned above), 2) low initial start–up costs because the baffles are simple in design and the mechanical refrigeration units are inexpensive, and 3) worry–free operation and elimination of typical maintenance problems encountered with liquid nitrogen baffle setups (e.g., unintentional emptying of liquid nitrogen dewars). With four baffles of this design on our apparatus, we accrue a yearly savings of ~$15,000 when compared to similar systems that use liquid nitrogen traps. These baffles are the subject of a manuscript we have submitted to Review of Scientific Instruments.