4.26μm sensor CO2

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

4.

26μm RCLED and a fast, low-power CO2 sensor


Audrey Nelson, Andrian Kouznetsov
GE Sensing, Telaire Products, 6860-B Cortona Drive, Goleta, CA-93117, USA

ABSTRACT

We have developed 4.26 μm room temperature resonant-cavity light-emitting diodes (RCLED) using the III-V material
system (InAs, InSb, and GaSb). The emission spectrum of RCLED closely matches the CO2 absorption profile. The
wavelength of the peak of the emission spectrum is stable with respect to wide variations in the temperature and the drive
current. Consequently we were able to develop a low cost RCLED based CO2 sensor operating over a wide temperature
ranges of +5 to + 45 degrees C and minus 20 to + 60 degrees C (depending on the IR detector used) without a TE-cooler.
These sensor devices consume very low power, at the level of mW, with a fast response time of 10 sec and better. They
can measure CO2 levels in the range 0-10,000ppm with an accuracy of better than +-5%.

Keywords: Mid-IR, RCLED, III-V materials, CO2 sensor, NDIR sensor, InAs/InAsSb

1. INTRODUCTION
The development of affordable and reliable NDIR CO2 sensors in the last decade has led to a growing number of
applications of these sensors in HVAC and other industries. GE Sensing, Telaire [1] division has long been one of the
leaders in development of low-cost NDIR technology and its applications. Traditional NDIR technology utilizes
broadband IR sources, such as lamps and micromachined blackbody emitters. The shortcomings of those systems are
instability and high power consumption.
Mid-IR LEDs developed in the last decades and currently available from a limited number of research and
development organizations suffer from drawbacks caused by instability of the radiation spectra due to temperature
variations. The NDIR sensor devices using these traditional Mid-IR LEDs often require a TE cooler, which in most cases
negates the advantages of using the LED.
The fundamental difficulty of Mid-IR semiconductor devices arises from the low energy (in the tenths of electron
volts) of Mid-IR photons, so the materials appropriate for this energy range must have a narrow band gap. Consequently
losses due to Auger recombination, where electron-hole pairs recombine and transfer their energy and momentum to
another carrier instead of emitting a photon, could become dominant especially at room temperatures and above. In
certain materials (for example InSb) energy splitting due to spin-orbit interactions becomes comparable to the band gap
itself, resulting in heavy intervalence band absorption. These fundamental limitations have prolonged Mid-IR room
temperature LED development for decades. Most of commercially available devices have unacceptable spectral
variations over a suitable temperature range.
This paper presents the latest technology development that has a goal of improving the design of a Mid-IR LED, making
it suitable for operation over an industrial temperature range without a TE cooler. Performance of this LED based CO2
sensor is also presented.

2. 4.26 μM RCLED

As pointed out above, typical hetero-structure Mid-IR LEDs exhibit shifts in radiation spectrum due to temperature
variation. We therefore put effort into the Resonant-Cavity LED (RCLED) design in order to stabilize the wavelength of
peak radiation intensity and overall radiation spectrum shape. The concept of the RCLED was first introduced and
developed by Dr. Schubert in 1990 [2] for Near-IR and visible wavelength range. There are many advantages of an

Light-Emitting Diodes: Materials, Devices, and Applications for Solid State Lighting XIII,
edited by Klaus P. Streubel, Heonsu Jeon, Li-Wei Tu, Proc. of SPIE Vol. 7231, 72310V
© 2009 SPIE · CCC code: 0277-786X/09/$18 · doi: 10.1117/12.809293

Proc. of SPIE Vol. 7231 72310V-1


RCLED compared with conventional LEDs, such as higher spectral purity, narrower angular distribution, higher
brightness and higher efficiency.
The emission line width of an RCLED is determined by the cavity quality factor Q, not by the thermal energy kT.
Therefore, one can optimize the emission line widths to match the absorption spectrum of a target gas. At the same time,
the RCLED has much better temperature stability compared with conventional LEDs because the wavelength shift with
temperature is determined by the thermal expansion coefficient of the optical cavity and not by the energy gap of the
active material.
So far most RCLED development has been in the visible and near IR region. A Mid-IR RCLED has certain challenges
with its material system. However, as we demonstrated for a typical device with a few micron thicknesses, the cavity
length fits the lowest orders of the optical modes for Mid-IR wavelengths. This promises a better overlap between the
resonant optical mode and the active region emission spectrum.

2.1 LED design and RCLED1


The first important issue of the LED design is to provide good carrier confinement so that most of electrons
recombine with holes in the active region and produce the desired photons, instead of being absorbed outside of the
active region. For the InAs/InAsSb material we use, the quantum well has shallow barrier such as 0.07~0.1 eV, which
causes heavy carrier overflow:

4 Nc 2 ΔEc 3
Joverflow = ( ) ( ) eBWDH (1)
3 π kT

~ (3-6) A/cm2
Fig 2.1.1 Shallow QW with carrier overflow
Where Nc is density of state, ∆Ec is barrier height, B is the bimolecular recombination coefficient, and WDH is the
thickness of active region.
Therefore in InAs/InAsSb system one quantum well (QW) will not be sufficient. Experimental data has shown that
the carrier escape from QW and recombination in cladding layer or substrate is dramatically reduced with a multiple
quantum well (MQW) design. Additionally in order to reduce electron escape from the active region, electron-blocking
layers are necessary. Implementation of MQW combined with electron-blocking layers helps to reduce carrier escape
and improve the efficiency of photon generation.

The second most important issue of LED design is improvement in the light extraction efficiency, because typically
the majority of the photons produced in the active region are lost. An increase in the light extraction efficiency can be
achieved by placing the active layers in a micro-cavity. This also helps to control the emission pattern. The LED top
surface and an internal distributed Bragg reflector (DBR) below the MQW form two cavity mirrors. Active layers
(MQW) need to be properly placed at the anti-node of the cavity where the strongest electro-magnetic field is located, as
shown on the figure below.

1
All the equation in this section can find in reference [3].

Proc. of SPIE Vol. 7231 72310V-2


Fig 2.1.2 Carton of RCLED structure

In order to allow radiation exiting from one mirror only (top mirror on the figure 2.1.2 in this case), the reflectivity
R1 of that mirror has to be much less than the reflectivity R2 of another mirror (DBR on the figure 2.1.2). In order to
minimize the self-absorption in the active region, one has to make sure that the following condition is fulfilled.

2ξαLactive < (1- R1) (2)

This means if the thickness of active region of Lactive ~ 600A, the absorption coefficient of α ~ 104cm-1, and the antinode
enhancement factor of ξ ~ 1.5, the R1 has to be < 82%.

Because the cavity quality factor Q defines the radiation spectrum width, we can calculate the proper reflectivity to
match the radiation spectrum width with the CO2 absorption lines using equations 3 and 4. For example, to match the
CO2 absorption spectrum line width of 0.25 micron the optimal Q = 17. If we wanted to have a narrower width of 0.2
microns then we can design Q = 21.3.
λ 4.26
Q=( )= = 21.3 (3)
δλ 0.2
Since also:

Lc
Q = 2π ( ) [- ln( R1R2 ) ]-1 (4)
λ
Where Lc is the cavity length and λ is the wavelength of the emitting light. We then can choose the proper value of R1
and R2 to achieve the desired spectrum line width. For example, with R1=0.3, R2=0.53, we have

Q = 2 π ∗ 2.5 ∗ [- ln( 0.3 * 0.53 ) ]-1 = 17

The emission pattern is directed mostly along the optical axis of the cavity. And the output radiation intensity is
enhanced about factor of 2-10. In this case, with ξ = 1.5, τ cav / τ ≈ 1, R1=0.3, R2=0.53, the emission rate enhancement is

τ cav ( R1 R2 )1/ 4 (1 − R1 )
G =ξ( )( )≈2 (5)
τ (1 − R1 R2 ) 2

Proc. of SPIE Vol. 7231 72310V-3


2.2 The RCLED radiation spectrum
In 2004, we successfully grew the RCLED structure using MBE. The spectral data on the figures below were measured
in arbitrary units using an FTIR spectrometer. The emission spectrum of the RCLED matches the CO2 absorption profile
very well, as seen in Figure 2.2.1 below. Further narrowing of the LED spectral distribution can be achieved by
increasing the reflectivity of the elements of the Resonant Cavity. The power output of the RCLED is in the order of 100
μW with drive current of 200 mA.

2.2 -
Ambient
1 .8
r12 Purge

0.6

0.2
3 3.5 4 4.5 5 5.5
Wavelength (1im)
Fig. 2.2.1. RCLED radiation spectrum as measured by FTIR spectrometer. Red line in under N2 purge in FTIR path and
blue line is at ambient condition.
Compared with a conventional LED, the spectral distribution of our RCLED is much purer. One of the major
advantages of this device is that there is no significant spectral shift over temperature range 0-50C or under different
driving currents. Refer to Figure 2.2.2 for LED response to driving current and Figure 2.2.3 for temperature response.
3-
2.5

0.5

0
3 3.5 4 4.5 5 5.5
Wavelength (p.m)
Fig. 2.2.2. RCLED radiation spectrum under different drive currents

Proc. of SPIE Vol. 7231 72310V-4


0
3 3.5 4 4.5 5 5.5
Wavelength (jim)
Fig. 2.2.3. RCLED radiation spectrum under different temperatures

3. FAST AND LOW-POWER NDIR CO 2 SENSOR USING RCLED

3.1 Typical commercially available low cost NDIR CO2 sensor


The goal of this project was to develop a quantum source that can be used in practical low cost NDIR applications,
including industrial environments with extended temperature range.
As was mentioned in the introduction section, typical commercially available NDIR CO2 sensors use an IR lamp or
a micromachined IR emitter as a source. These sensors are robust and reliable but in any practical application the power
consumption can be in the hundreds of mW range.
The outline of NDIR CO2 sensor, commercially available from GE Sensing, Telaire is presented on the Figure 3.1.1.

Custom Designed
Infrared Filters
Reference Target Gas Dual Beam
Diffusion Membranes Micro-Machined
Thermopile Detector

Incandescent 1.25"
Infrared Source Patented Waveguide

Microprocessor

Fig. 3.1.1. Typical commercially available low cost NDIR CO2 sensor
The source is an incandescent IR lamp with broadband emission; a narrow band filter is mounted on the detector. By
modulating the lamp power we can reduce any background radiation bias, detector drift or electronic drift. The gas-

Proc. of SPIE Vol. 7231 72310V-5


sampling chamber is also a wave-guide, designed using optical ray tracing tools to minimize optical loss. By using a dual
beam detector, the system is able to correct for any light source and optical system degradation bias. However for many
applications lower cost single beam designs are used.

3.2 Low power and fast response time CO2 sensor using RCLED
The primary goals of our RCLED development are the use of the LED in a simple and robust sensor design (as
Figure 3.1.1) and the reduction in power consumption by orders of magnitude, compared to existing commercial
products.

We have developed and tested multiple prototypes of a CO2 sensor in order to understand the performance of the
LED in a practical sensor system. The difference between these prototypes was primarily in the detector and associated
preamplifier electronics. We tested the prototypes using a typical absorption scheme with both pyroelectric and
thermopile IR detectors. The optical outline of the prototypes is very simple and similar to the one on the Figure 3.1.1
with the only difference being the use of the RCLED instead of the lamp. Alternatively the same RCLED can be used in
a photo-acoustic scheme, which currently is being developed, but the results are not ready for publication yet. The
prototypes were built to measure CO2 concentration in the range from 0 to 1% CO2.
Using the fast pyroelectric detector one can build a rather economical sensor system. The fast pyroelectric detector
that we used has a 2 msec rise time. This sensor prototype has a 10 second response time to CO2 concentration range,
5% relative sensitivity and operates at the temperature range of +5 to +45 degrees C. The RCLED was driven by 100 mA
current. The temperature range is limited by the pyroelectric detector. The sensor module consumes on average 13 mW
power from 6V power supply, which was limited not as much by LED as by electronic design.
The sensitivity, response time and power consumption are related to each other. For example one can have sensitivity of
5% with 10s response time or have 2.5 % sensitivity and 10s response time, but power consumption will double. Or one
can have better sensitivity with same power consumption but slower response time.

12000

10000

8000

6000
U 4000
2000

0
0 500 1000 1500
Sample
Fig. 3.2.1. Sensor output vs. CO2 concentration profile with 5% sensitivity and ~10s response time

Proc. of SPIE Vol. 7231 72310V-6


12000 -
10000

8000

6000

U 4000
2000

0
0 500 1000 1500
Sample
Fig. 3.2.2. Sensor output vs. CO2 concentration profile with 2% sensitivity and ~20s response time
Sensors with RCLEDs have demonstrated very good stability over time. The Figure 3.2.3 presents the results of
CO2 concentration measurement in pure Nitrogen. Figure 3.2.3 presents 1% CO2 concentration data. The data was taken
from the same sensor with 5-month interval in time. Between these two measurements the sensor was continuously
operating. The data on figure 3.2.3 have an error from the desired 1% (10,000 ppm) CO2, which was caused by initial
calibration error. These results demonstrate exceptional long-term stability of RCLED.

200 -
150 -Dec-08 -
100
50

C\1
0
0
-100
-150
-200
0 1500 3000 4500
Samples
Fig 3.2.3. RCLED CO2 sensor zero stability over long period of time

Proc. of SPIE Vol. 7231 72310V-7


10500 -

10300

:10100
9900-
9700

9500
0
- 8 1 .2008
1152009

50 100 150 200 250 300


Sam pies
Fig 3.2.4. RCLED CO2 sensor stability at 10,000 ppm over long period of time

In order to achieve a wider temperature range, we replaced the pyroelectric detector with fast thermopile detector.
The detector had a 15 msec response times with reasonable sensitivity. Figure 3.2.5. shows very good linear temperature
response of the CO2 sensor with RCLED and thermopile detector. The blue line is the sensor optical signal intensity
profile in relative units over the temperature range –10C to 60C when sensor measures pure Nitrogen. The purple line is
the same profile with the sensor measuring 2000 ppm of CO2 in Nitrogen. This data demonstrates that the RCLED
output is changing linearly over a wide temperature range and that it can be used to operate at temperatures up to 60
degrees C without TE cooling.

2.2 -
CO2 N2
C 02 2000P PM
1.8

1.4

0.6

0.2
0 2500 5000 7500
Time
Fig. 3.2.5. Sensor temperature profile –10 to 60C with fast thermopile detector.

Proc. of SPIE Vol. 7231 72310V-8


4. CONCLUSION

In conclusion, we have developed the Mid-IR RCLED. The Mid-IR RCLED has demonstrated exceptional stability
of the peak wavelength over current and temperature and stability of the output power over time. Prototypes of CO2
sensor based on this LED were developed and confirmed that it could be used in practical applications including
industrial environment, with a potential of achieving power consumption in the order of 100 μW. It can be used in
practical applications including industrial environment, without a need for TE (or any other type of) cooling.

REFERENCES
[1]
Telaire: www.telaire.com.
[2]
Schubert E. F., Wang Y. -H., Cho A.Y., Tu L. W., and Zydzik G. J. “Resonant cavity light emitting diode” Appl.
Phys. Lett. 60, 921 (1992)
[3]
Schubert E. F., [The light emitting diode], Cambridge University Press, Cambridge, (2003).

Proc. of SPIE Vol. 7231 72310V-9

You might also like