1887
banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
Real-time material quality prediction, fault detection, and contamination control in high electron mobility transistor metalorganic chemical vapor deposition process using in situ chemical sensing
Rent:
Rent this article for
USD
10.1116/1.2006110
/content/avs/journal/jvstb/23/5/10.1116/1.2006110
http://aip.metastore.ingenta.com/content/avs/journal/jvstb/23/5/10.1116/1.2006110
View: Figures

Figures

Image of FIG. 1.
FIG. 1.

Real-time prediction of GaN PL band-edge intensity during growth based on in situ mass spectrometry measurement of methane∕ethane byproducts ratio (precision 4.8%). The four runs shown here correspond to the same four runs presented in our previous publication on crystal quality prediction, (see Ref. 20) where it was shown that the same methane∕ethane ratio metric predicts GaN epilayer crystal quality to 3.5% verified by postprocess XRD. Moreover, the correlation is such that by going to lower methane∕ethane ratio we improve material quality as seen in both XRD and PL.

Image of FIG. 2.
FIG. 2.

Correlation of the gas phase residual level within the pregrowth reactor to the GaN PL band-edge intensity measured postprocess. Better quality material (as seen in higher band-edge intensity) was obtained when the pregrowth residual level was lower with the correlation accurate to 6.6% here.

Image of FIG. 3.
FIG. 3.

(a) Correlation of the gas phase residual impurity level, measured at , within the pregrowth reactor to the GaN PL band-edge intensity measured postprocess. Better quality material (as seen in higher band-edge intensity) was obtained with higher (not lower) impurity level detected prior to growth with the correlation accurate to 10% here. (b) Further study is required to determine the exact identity of the unknown specie at and its effect on the process. However, the objective of the current publication was to identify working metrics for immediate implementation of process fault detection and material quality prediction. A sample full spectra taken from the pregrowth reactor is shown as a reference. Mass resolution capability was limited to , and for optimum response time required for real-time process control only a selected set of amus were monitored at 2, 13, 17, 18, 26, 27, 28, 32, and .

Image of FIG. 4.
FIG. 4.

Correlation of the gas phase residual level within the pregrowth reactor to the GaN PL deep-level intensity measured postprocess. Better quality material (as seen in lower deep-level intensity) was obtained when the pregrowth residual level was lower with the correlation accurate to 19% here. Although the correlation here is less precise compared to the case of PL band-edge intensity, this offers an important means to predict and possibly control GaN deep-level concurrent with the band-edge intensity, because the two parameters put together are critical as a measure of GaN material quality.

Image of FIG. 5.
FIG. 5.

Example of real-time fault detection based on clear chemical signatures for process and equipment faults which potentially lead to unacceptable product quality. (a) Desorption of vapor from the unconditioned walls of a brand new liner as soon as the temperature ramp begins, followed by desorption of impurity coating from a dirty susceptor in the form of during the high temperature purge. (b) Such impurity evolutions were not observed with a well conditioned liner and a clean susceptor. These kinds of real-time indications as in (a) clearly correlated to unacceptable material quality by PL, and they indicated a need for measures to correct the root cause, such as through extended pregrowth contamination control and replacement of corresponding tool parts (e.g., susceptor). The cause of signal loss at in (a) and in (b) is an intentional feature (used for electron multiplier gain study) and may be ignored for the purposes of analyses here.

Image of FIG. 6.
FIG. 6.

Example of real-time detection of an equipment excursion in the middle of process that led to unacceptable material quality. Failure of an MFC was immediately detected by the mass spectrometry during GaN epilayer growth from the real-time signatures indicating large disturbances in the concentration within the reactor. Segment A indicates a manual test where the isolation valve between the mass spectrometer and the reactor was closed for a period of . Stable signal (still remaining high due to the trapped volume within the sampling inlet of the sensor) during this period as well as the time scale for each sensing scan confirmed that the disturbances were real physical effects and not an electrical noise. Segments B and C indicate a series of MFC flow variation tests where the set point for the MFC flow rate was altered between a high value (80% for B) and a low value (40% for C). Results of the test indicate that the MFC failed to perform properly for low flow rate set point conditions. The MFC in question was replaced immediately and both the sensor signals as well as the future product materials themselves indicated that the fault was successfully resolved.

Image of FIG. 7.
FIG. 7.

Mass spectrometry sensing during a run where the reaction byproduct ( and ) levels were significantly lower than the usual. This was correlated to the little growth of product film seen postprocess, which led to examination of a limited number of possible sources for the fault. Once the root cause was identified to be untimely precursor source depletion, the precursor source bottle was immediately replaced. The segment labeled “EM off” indicates a test where the mass spectrometer electron multiplier was turned off for and then turned back on to confirm that it was functioning properly.

Loading

Article metrics loading...

/content/avs/journal/jvstb/23/5/10.1116/1.2006110
2005-08-15
2014-04-23
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Real-time material quality prediction, fault detection, and contamination control in AlGaN∕GaN high electron mobility transistor metalorganic chemical vapor deposition process using in situ chemical sensing
http://aip.metastore.ingenta.com/content/avs/journal/jvstb/23/5/10.1116/1.2006110
10.1116/1.2006110
SEARCH_EXPAND_ITEM