for medical diagnostics with 1.6-ptm distributed-feedback semiconductor diode lasers. David E. Cooper, Ramon U. Martinelli, Clinton B. Carlisle, Haris Riris,.
Measurement of
1 2 CO2 :13 CO
2
ratios
for medical diagnostics with 1.6-ptm
distributed-feedback semiconductor diode lasers David E. Cooper, Ramon U. Martinelli, Clinton B. Carlisle, Haris Riris, D. B. Bour, and R. J. Menna
We have observed some of the absorption lines from the molecules 2CO2 and 3CO2 in the 1.6-pm spectral region with the use of specially fabricated single-mode InGaAsP distributed-feedback semiconductor diode lasers. Using a 23.6-m-long multipass absorption cell in combination with radio-frequency modulation and detection techniques, we measured the 2 CO2 :13CO2 isotopic ratio of two specificlines at 6253.73 and 6253.90 cm-'
human breath.
with sufficient precision for diagnostic medical tests that analyze CO 2 on
The widespread use of tunable III-V compound semiconductor diode lasers in optical communications and consumer electronics has made them the subject of intensive development for over 20 years. As a result of this effort, diode lasers emitting at selected wavelengths in the 0.8-1.55-[um region are technically advanced and commercially available. One emerging application of these devices is trace-gas analysis and near-infrared spectroscopy. Many molecules, including common environmental pollutants and toxic chemicals, have absorption bands between 1 and 2 [um. These bands are strong enough to allow detection at or below the parts-per-million level with a near-infrared diode laser. Although these sensitivities are several orders of magnitude lower than what is achievable in the mid-infrared range with lead salt diode lasers that monitor fundamental-band absorptions, they are large enough for a number of important industrial and medical applications. One particularly interesting application of III-V diode lasers operating at 1.6 m is breath analysis through the use of 3 C-labeled substances. 13Cis a stable, benign isotope with a natural abundance of When this research was performed, D. E. Cooper, C. B. Carlisle, and H. Riris were with SRI International, 333 Ravenswood Avenue, Menlo Park, California 94024; R. U. Martinelli, D. B. Bour,
and R. J. Menna were with the David Sarnoff Research Center, CN 5300, Princeton, New Jersey 08543-5300. D. B. Bour is now with Xerox PARC, 3333 Coyote Hill Road, Palo Alto, California 94304. Received 4 November 1992.
0003-6935/93/336727-05$06.00/0. C 1993 Optical Society of America.
approximately 1.1%, and this abundance fluctuates by approximately 5%. The traditionally used isotope 14C is radioactive, and it presents health and handling problems. In CO2 diagnostic testing the procedure is to administer a 3C-labeled substance that is broken down by a specific enzyme in the human body to produce 3 C02, which is then exhaled on the breath. Samples of breath are collected over a period of several hours after ingestion of the labeled substance, and each sample is analyzed to determine the 12CO2: 13CO 2 ratio. By using this method with different labeled compounds, a physician can screen patients for glucose utilization, pancreatic function, intestinal bacterial overgrowth, liver function, and H. pylori infections of the digestive tract.' The techniques for monitoring 13Cinclude isotoperatio mass spectrometry, nuclear magnetic resonance spectrometry, and infrared absorption spectroscopy. Infrared approaches as described above include modulation spectroscopy with lead salt diode lasers to detect the C02 fundamental-band absorptions in the 4.3-Rm region,2 and infrared heterodyne ratiometry using an incoherent light source to detect absorption from the same C02 bands.3 This paper discusses the spectroscopic method of detecting 13CO2 and 12C02 that utilizes their combination absorption bands in the 1.6-vumspectral region. This is an attractive approach to isotope-ratio measurements because the combination bands are strong enough for detection at the sensitivity levels required for medical diagnostic tests, and the diode lasers used are inexpensive, highly reliable, single-mode devices. Furthermore, the near-room-temperature operation 20 November 1993 / Vol. 32, No. 33 / APPLIED OPTICS
6727
of the laser and detector results in a much simpler and compact instrument than is possible with the other techniques. In the following sections we discuss our experimental apparatus, the properties of the single-mode laser sources, the near-infrared spectrum of C02, and measurements of C02 on actual
human breath samples. The experimental apparatus is illustrated schematically in Fig. 1. The 1.6-pm single-mode laser, described in detail below, is mounted on a thermoelectrically cooled heat sink. A commercial diode-laser controller supplies the laser bias current, and it also
controls the heat-sink temperature (0-40 00). A function generator supplies a 1-kHz current ramp to the laser diode that repetitively sweeps the laser frequency (wavelength) through a spectral interval of approximately 10 GHz, encompassing several absorption lines. The collimated laser beam enters a multipass absorption cell of the Herriott design. Basically, this cell consists of two identical 7.6-cm-diameter spherical mirrors mounted at the ends of a Pyrex vacuum tube. The mirrors have a 46-cm focal length and are separated by a variable distance that determines the total number of passes through the cell. One mirror is fixed in position and provides optical access to the cell through a small clearance hole near the periphery of the mirror. The diode laser beam enters the cell through a vacuum window centered on the clearance hole. At specific values of the mirror separation, the cell is re-entrant: at a mirror separation of 73.6 cm, the beam is reflected back and forth between the two mirrors 32 times before exiting through the clearance hole at a small angle with respect to the entrance beam. The total pathlength is 23.6 m. The total cell volume is approximately 3.3 L. The exit beam impinges onto a fast InGaAs photodiode, in which it is square-law detected. To measure the C02 absorption lines with high sensitivity, we modulate the laser current at radio frequencies, and we detect the appropriate rf Fourier component in the detector photocurrent. We have used two specific modulation techniques in our experiments. In the first, known as two-tone frequency modulation spectroscopy (TTFMS), the laser diode is driven at two rf frequencies: 195 and 205 MHz.
Phase-sensitive detection occurs at the 10-MHz intermediate frequency. In the second technique, known as wavelength modulation spectroscopy (WMS), we modulate the diode laser at a single radio frequency, 5 MHz, and detect signals at the second harmonic of the applied rf, 10 MHz. Both modulation techniques generate signals from the molecular-absorption lines that have a second-derivative character. The signal amplitudes are proportional to the power of the laser beam and to its absorption traversing the multipass cell.
TTFMS and WMS use the same 10-MHz synchronous-detection circuit. It consists of a 10-kHzbandpass crystal filter with a 10-MHz center frequency, a pair of low-noise rf amplifiers separated by a variable attenuator to allow adjustments of the total gain, and a double-balanced rf mixer referenced to the rf source that drives the diode laser. The mixer output is a 1-kHz repetitive waveform that contains the TTFMS or the WMS line spectra of the gas sample. The output is filtered with a 10-kHz lowpass filter and 256 waveforms were averaged together using two 8-bit digital oscilloscopes(Tektronix 2403A) as described below. The signal amplitudes were measured using the peak-detection function of the scopes. The effective-detection bandwidth in these experiments was 1 Hz. Because the 12CO2 signal is 20 times greater than the 3CO2 signal, we used two different digital oscilloscopes to measure the 12CO2 and 13CO2 signals. Each scope gain was adjusted to give a full-scale reading for its respective signal, permitting full 8-bit, or 0.4%, accuracy in the amplitude measurements. To check the line positions accurately, a removable mirror inserted into the laser beam reflects it into an interferometric wavemeter, which is accurate to 0.01 cm-' (0.003 nm at 1.6 pum). The laser is statically tuned (no ramp or rf currents) to the absorption-line peak, and its wavelength is measured. We use a calibrated etalon to determine the spacings among the lines. An important consideration in the use of highfrequency modulation techniques for sensitive measurements is the minimum detectable-gas concentration in a given experiment. Essentially, the highest sensitivity achievable in an optical-absorption experiment is the so-called quantum-noise limit. Light of power P incident on a photodetector generates a photocurrent
i given by
i,= (e/hv)P, where 1qis the photodetector
Multipass(Herrio) absorption cil
Fig. 1. Experimental setup showing the laser, the absorption cell, the detector, and the electronics. A removable mirror directs the beam into a wavemeter for determining the lasing wavelength in
situ. 6728
APPLIED OPTICS / Vol. 32, No. 33 / 20 November 1993
(1)
quantum efficiency, e is
the electronic charge, h is Planck's constant, and v is the frequency of the light. Associated with this signal photocurrent is a shot-noise photocurrent given by ij
=
(2eiB)1/ 2,
(2)
where B is the detection bandwidth. The detection signal-to-noise ratio (SNR) is the ratio of these two
photocurrents, and is given by SNR = i/in = [/2hv)P/B]
12
1 /2 .
(3)
As an example of what this SNR means, consider a 0.5-mW, 1.6-pm diode laser beam incident on a photodetector of quantum efficiency 0.8, a situation realized in the present experiments. Assuming a detection bandwidth of 1 Hz, the above expression gives SNR = 4 x 107. This allows an absorption
approximately as small as 1/SNR or 2.5 x 10-8 to be detected. Given a 100-Torr breath sample nominally containing 4% total C02 in a 23.6-m absorption cell, this sensitivity allows detection of the stronger 1.6-pum 3CO2 absorption lines with a SNR of nearly 6 x 104 and a corresponding measurement precision of 0.002%. A more practical instrument working sensitivity of 10-6 would allow a SNR of about 1400, corresponding to a measurement precision better than 0.1%. This measurement precision is well above the natural 5% variation of 13CO2 on human breath. The lasers used in this study were strained-layer,
multiquantum-well, separate-confinement-heterostructure, distributed-feedback, InGaAs-InGaAsP diode lasers. The growth, fabrication, and optoelectronic characteristics of these devices as FabryPerot diode lasers has been discussed elsewhere.4 The present lasers have a first-order Bragg grating incorporated into their structure. The grating provides the resonant distributed feedback that supports laser action. A high-reflection and a low-reflection coating on the two output facets create an asymmetry that leads to a single-mode output spectrum. 5 Detailed measurements of the laser output from 5 to 28 °C show the device operating in a single spatial and longitudinal mode at an output power of 5 mW. The measured side-mode-suppression ratio over the same temperature interval is at least -30 dB. We estimate the linewidth to be less than 50 MHz at 5-mW output power. To access the molecular lines of interest in this work, this laser was cooled to 7.8 0C and was biased at approximately 140 mA. Under these conditions the wavelength was in the neighborhood of 1.599 pm, or 6254 cm-'. The output power characteristics of the laser, taken at the output of the collimating lens and at 7.8 C, is shown in Fig. 2. The threshold current is 92 mA, and the differential slope efficiency near threshold is 23%. At 140 mA, the output power is 7.8 mW. Above 165 mA, the slope efficiency de-
creases, and the output power reaches its maximum
10 0
8
IL
a,
6
I.
04
0
0o
4 2 0 - 'W I I I I I
80
100
rate of -0.57 cm-1/K (0.15nm/K) and with current at a rate of approximately -900 MHz/mA. The frequency interval covered by the tuning curves is 1.8 cm-1, or approximately 53 GHz, which was sufficient to monitor the C02 lines of interest. Figure 4 shows the absorption bands of 12C02 and 13CO2 in the spectral region of 6200-6270 cm-1
I
140
I
160
I
II
180
200
II
,
220
Diode Current, Id(mA)
Fig. 2. Power-current characteristics at 7.8 C of the distributedfeedback laser.
(1.59-1.61 m) taken from the HITRAN database of molecular absorption lines in the atmosphere (1986 edition).6 The strongest 12C02 bands result from transitions between the ground state (00001) and the v1v2lv 3r = 30013 level and have strengths in the range of 10-24-10-23cm/molecule. The 13CO2lines result from transitions between the ground state and the 30012 level and have strengths in the range of 10-26-10-25 cm/molecule. In identifying lines and comparing line strengths measured in the laboratory, it should be remembered that in the HITRAN data, the relative natural abundance of 13CO2is factored into the line strength. Therefore, the HITRAN data give the correct line-strength ratios only for a sample containing natural C02. Figure 5 shows the 12CO2and 13CO2spectra taken from a 109-Torr breath sample. A TTFMS spectrum is shown in Fig. 5a, and a WMS spectrum is shown in Fig. 5b; each spectrum is taken with the modulation index adjusted to give a maximum signal. Each spectrum represents the average of 256 waveforms corresponding to a detection bandwidth of about 1 Hz. Both spectra have the same characteristic line shape approximating the second derivative of the absorption line. The position of the 12CO2line is 6255
. . . .. .. .I . . .
-900MHz/mA 1.5987
T= 700-C So le T=7 7.8
E
0
-900 MHz/mA 71,1 5
