You are on page 1of 12

Novel spectrometer architecture for photoluminescence measurements of

semiconductor thin films grown by molecular beam epitaxy


Giancarlo Sabetta
February 25, 2015

Abstract

Photoluminescence measurements are crucial to the study and characterization of thin


film semiconductor material systems. As semiconductor devices that operate in the near and mid
infrared are gaining popularity for their commercial and noncommercial uses, measurements into
such wavelengths are becoming increasingly necessary. Specifically, the development of a
photoluminescence setup capable of measurement into these wavelengths expands upon previous
widespread capabilities of measurement in the visible and shorter wavelengths. In this work the
researchers posed one method to measure at these longer wavelengths: a low cost Arduino
microcontroller used as a hardware controller for a Spex double monochromator and a Jobin
Yvon photodiode. Custom code and wiring was used to interface the devices together and the
system was tested for its accuracy and precision using samples of known spectra. The system
was shown to be a successful method for photoluminescence measurements into the near infrared
with results from the system in good agreement with the expected spectras of tested materials.
Additionally, a low temperature measurement system was added to aid in the detection of poor
radiative samples.
Introduction
Semiconductor devices operating in the infrared have many scientific, medical,
environmental, industrial, commercial and military applications. For example, infrared devices
have been used in everything from TV remotes to missile guidance systems and are becoming
increasingly useful in the detection of gasses in the atmosphere. The study and characterization
of semiconductor material systems that emit in the infrared are therefore vitally important to the
improvement and continued development of these devices (Garcia, 2013).
One method for the fabrication of such devices is molecular beam epitaxy (MBE), a
process in which thin epitaxial layers of material are deposited onto a substrate under ultra high
vacuum. These layers then bond with each other to form a crystal in a periodic lattice structure.
Using MBE, crystals can be grown with precise compositions, allowing for the study of varying
material systems on device performance.
After samples are grown by MBE, they are subjected to a suite of characterization
techniques in order analyze the various properties of the crystal. One such vitally important
characterization technique is photoluminescence. Broadly, photoluminescence (PL) is the
spontaneous emission of electromagnetic radiation from a material when excited by an optical

source (Gfroerer, 2006). Specifically, after an optical source excites valence bound electrons into
the conduction band of a material, the electrons relax down to their ground state and generate a
photon equal in energy to the electrons transition. The energy of these photons is then used to
determine the band structure of a given a material, including bandgap(s), heavy/light hole
transitions, etc. (Shen, 2013). PL measurements can also be used to provide information on
crystal quality; specifically, PL is used to highlight crystalline defects within the material. Due to
its ease of use and non destructive procedure, PL measurements are an attractive characterization
technique in situations where material availability is limited, the material is fragile, or has an
electrically resistive surface (and as such contacting the material is difficult). For samples that do
not produce strong emission, cooling the material to low temperatures (<77K) is often employed
to increase the strength of the signal (Gfroerer, 2006).
To conduct PL measurements, an optical excitation source (such as a laser), is needed to
pump electrons into the conduction band, and a detector is needed to examine the resultant
emitted light. Crucially, the energy of the laser must be greater than the bandgap of the material,
(so that excited electrons are able to reach the conduction band from the valence band The
range of the detector must be in the expected emission of the material.
One type of detector apparatus is a combination of a photodiode and a monochromator.
The monochromator splits light into discrete wavelengths and the intensity is then read by the
photodiode. In this way, the photodiode does not measure the wavelength of light it is measuring,
only its intensity, while the monochromator is not aware of the intensity of light passing through
it, only its wavelength. By interfacing both of these devices, a spectra showing wavelength
versus intensity can be obtained from the samples emitted light. Most photodiodes return a
voltage output linearly proportional to the intensity of incident to the sensor, while many
monochromators are controlled by some variant of a serial connection. Many spectrometers that
employ the photodiode/monochromator method of detection are sold either together or are able
to interface with larger, proprietary control hardware.
Within the last few years, new advancements in microcontrollers have made possible
robust, inexpensive computing power. One such device is the Arduino Mega, a type of
microcontroller board that runs on an ATmega 2560 processor at 16Mhz. The board is used all
over the world from projects ranging from blinking a light bulb to autonomously driving cars.
The boards fast clock speed and robust processor make it ideal for the collection of

photoluminescence data, which needs to be rapidly obtained and processed. Typically


spectrometer controllers cost huge sums of money and come with proprietary software, but by
using an inexpensive Arduino board open source code can be leveraged to take measurements
that are equal in quality to those of higher priced solutions. Additionally, Arduino is expandable,
and due to the nature of its architecture, can be molded to fit many different types of devices.
Methods and Materials
A Spex 1680B double monochromator was interfaced with a photodiode and lock-in pre
amplifier to reduce signal noise using an Arduino Mega 2560 in order to take spectral
measurements of MBE grown semiconductor materials. The monochromator had two gratings
blazed at 630nm with 1200 grooves per millimeter. The output from the Jobin Yvon germanium
solid state photodiode was fed to a lock-in preamplifier, which, in connection with a mechanical
chopper on the excitation source, regulated the signal and reduced noise. The excitation source
was an Omnichrome Series 74 HeCd Helium Cadmium Laser operating at about 325nm. For
measurements conducted at low temperature, a Janis Research cryostat was used.
Spectrometer hardware setup:
The HeCd laser beam passed through an iris, a focusing lens and a mechanical chopper
set at 400Hz, before reaching and exciting the sample. The samples resultant radiation was then
captured by two collection lenses and focused into the double monochromator. The photon flux
from the output of the monochromator was then read by a Ge photodiode. This voltage value was
first sent to a lock-in amplifier, which subtracted any background radiation from the signal using
the mechanical chopper. Care was taken not to saturate the detector or exceed the 1V output from
the lock-in amplifier. The scaled 0-1V output from the lock-in was read into an analog input pin
on the Arduino with a resolution of ~1mV. The Arduino was capable of 10,000 measurements per
second, and the limiting efficiency of the system was the speed at which the monochromator
could tune to a wavelength, not the Arduinos signal acquisition time. Two digital output pins on
the Arduino were wired to the Step and Direction pins on the monochromators external input
connection port. The manipulation of voltages to these ports controlled the direction and step of
the monochromator in 0.02nm increments. The grounds were connected and care was taken to
avoid ground loops.

Fig. 1 The layout of the hardware setup. The blue line is the beam path of the excitation laser,
while the green line is the beam path of the samples emission. Not pictured: Lock-in pre
amplifier, Ge detector.
Description of system control and data processing chain:
Two programs ran in parallel to support the operation of the spectrometer. The first ran on
the Arduino and handled raw data processing while the second ran on a laptop computer and
handled the data analysis and display. The Arduino based code accepted serial commands from
the computer. These commands instructed the Arduino to run a specific program. For example,
one program allows the user to measure light intensity from a given starting wavelength to an
ending wavelength and then returns a graph of emission intensity versus emission wavelength for
that range. Once a program was selected, the Arduino would pass raw data points back through
the serial line to be analyzed on the laptop computer. The program on the laptop was written in
MatLab and used the MatLab graphical user interface builder to create a user interface for
controlling the spectrometer. Both the Arduino and computer based programs communicated at
9600bps, and it was found to be sufficient for the data transfer. The Arduino based program
utilized the Arduinos onboard persistent EEPROM memory to keep track of the

monochromators current position. This was done so that the Arduino would maintain the most
accurate value of the monochromators position in case of a power failure or other critical
failure. A physical display readout on the monochromator was used to calibrate the Arduinos
EEPROM with the location of the monochromator before each measurement to ensure the most
accurate results. The position data was stored, read and written in raw motor steps. This has the
advantage of allowing the user to configure a single constant to change monochromator gratings.
The system was calibrated before each test to ensure precise results. The graphical user interface
performed 4 major functions: scanned, read intensity, moved the monochromator to a certain
wavelength and calibrated the monochromator.

Fig. 2 The graphical user interface located on the operating laptop computer. It provided a single
interface for users to calibrate the monochromator, retrieve and set the monochromators position,
measure the intensity of light being received by the photodiode, and scan the spectra of a sample
over a given wavelength interval. The preview window could be expanded for further
investigating and saved to a text file for import into a more robust plotting software.

Description of sample characterization


Samples were placed on a 3 axis stage and excited with the aforementioned HeCd laser.
Their radiative emission was focused through two collection lenses into the monochromator via
manipulation of the stage. When conducting experiments under low temperature, the same
methodology was used with the samples placed in a cryostat.
Results
Photoluminescence measurements were used to test the effectiveness of the methodology and
validate the use of an Arduino as a spectrometer controller. The result of the research was the
successful completion of an apparatus that will continue to serve the lab as they characterize
samples that emit in the deep infrared.

Fig. 3 PL of a red diode laser, with a manufactured operating wavelength of 635nm. The peak of
the luminescence occurred at 636.5nm and the full width at half maximum (FWHM) is 4nm. The
intensity has been normalized to one.

Fig. 4 PL a GaAs substrate. The peak of the luminescence occurs at 885nm and the FWHM was
72nm. There was no observation of secondary peaks in the spectra, although small lumps were
present to the right of the peak. The intensity was normalized to one.

Discussion
A photoluminescence spectroscopy setup was successfully developed that was able to
conduct spectral measurements into infrared wavelengths. The emission wavelength of the red
diode laser (636.5nm) observed in figure 1 was in very good agreement with the manufacturers
classification (635nm). The 0.3% difference between the expected and observed value is well
within the margin of error introduced by impurities in the the crystals. The very sharp peak, 4nm
width at half maximum observed in figure 1 is indicative of a diode laser and confirms that the

spectrometer and photodiode accurately worked in tandem to resolve spectral measurements at a


very high sample rate. In figure 2, the spectra of GaAs is much more broad. This is consistent
with bulk GaAs, however the non-symmetric nature of the peak is not. The reasons for the
shallow slant towards the trailing edge of the peak may be due to the age of the sample therefore
suggesting its deterioration, or possibly that the current grating in use was nearing the end of its
usable measurement range. The observed emission peak of 885 nm, is ~10nm away from the
theoretical peak of GaAs emission at 873 nm (Braunstein, 1959). This difference can be
attributed to various growth parameters, such as the accidental inclusion of selenium or indium
during fabrication. The small humps seen on the right of the peak may indicate deep level
transitions, which further supports the hypothesis that the sample was deteriorating or possessed
impurities.
The two programs that were written have been made available on GitHub at
https://github.com/kotran/Tamargo_Group_Spectrometer. The code is free to use and distribute
and is compatible with not only the Spex 1680 monochromator but all other monochromators in
that family. Moving forward the program and setup will be modified to include an option for
integration timing, a method of spectroscopy that increases dwell time on a specific wavelength.
This allows the detector to more accurately read the intensity of light and increase both accuracy
and significance. Additionally, tests comparing the system described herein versus a
commercially available spectrometer will be conducted to further confirm the validity of the
results.
Acknowledgements
I would like to thank Prof. Maria Tamargo for allowing me the opportunity to work and
learn in the molecular beam epitaxy lab at the City College of New York and for sharing her time
to teach me about MBE and formally introduce me into the materials science research
community. I would also like to thank Thor Garcia, Vasilios Deligiannakis, and Joel De Jesus for
their support and guidance in my research. Additionally, thank you to Ms. Schmitz for helping
me throughout my years in science research.

10

References
Arthur, J. R. (2002). Molecular beam epitaxy. Surface Science, 500(1-3), 189-217.
doi:10.1016/s0039-6028(01)01525-4
Braunstein, R. (1959). Intervalence band transitions in gallium arsenide. Journal of Physics and
Chemistry of Solids, 8, 280-282. doi:10.1016/0022-3697(59)90337-3
Garcia, T. A., Hong, S., Tamargo, M., Jesus, J. D., Deligiannakis, V., Ravikumar, A., . . . Shen, A.
(2013). Improved electrical properties and crystalline quality of IIVI heterostructures for
quantum cascade lasers. J. Vac. Sci. Technol. B Journal of Vacuum Science & Technology
B: Microelectronics and Nanometer Structures, 31(3). doi:10.1116/1.4803837
Gfroerer, T. H. (2006). Photoluminescence in Analysis of Surfaces and Interfaces. Encyclopedia
of Analytical Chemistry Applications, Theory and Instrumentation.
doi:10.1002/9780470027318.a2510
Shen, A., Ravikumar, A. P., Chen, G., Zhao, K., Alfaro-Martinez, A., Garcia, T., . . . Gmachl, C.
(2013). MBE growth of ZnCdSe/ZnCdMgSe quantum-well infrared photodetectors. J.
Vac. Sci. Technol. B Journal of Vacuum Science & Technology B: Microelectronics and
Nanometer Structures, 31(3). doi:10.1116/1.4794383

11

You might also like