High-speed vertical-cavity surface-emittinglaser (VCSEL) absorption spectroscopyof ammonia (NH3) near 1.54 µm1 Institute of Chemical Engineering, Vienna University of Technology, Getreidemarkt 9/166, 1060 Vienna,
Austria2 Walter Schottky Institut, Technische Universität München, Am Coulombwall, 85748 Garching, Germany3 Vertilas GmbH, Karl-Richter Str. 4, 80939 München, Germany
ABSTRACT A single-frequency VCSEL has been used for thefirst time for high-resolution spectroscopy near 1.5 µm. Theincorporated buried-tunnel-junction technology enabled the re-alization of a long-wavelength InGaAlAs/InP VCSEL with lowthreshold current (0.925 mA), high output powers (0.576 mW)and low series resistance (60 Ω). The high-speed tuning ca-pability of the long-wavelength VCSEL was investigated andused to conduct high-speed absorption spectroscopy. The peaktuning speed was measured to be 3.4 cm−1/µs and a 4.5-cm−1-wide NH3 spectrum was recorded in 2 µs. The VCSEL wasused to measure highly resolved low-pressure spectra for pres-sures ranging from 9.6 mbar to 1 bar. The measured Doppler-broadened linewidth of 0.02 cm−1 agrees within 3% with thetheoretical calculations. The availability and various advan-tages of 1.3–2-µm single-frequency VCSELs as compared toedge-emitting diode lasers, such as a large current tuning rangeeven at very high tuning frequencies, and low production costs,should significantly expand the application fields for near-infrared laser gas sensors.
PACS 42.62.Fi; 82.80.Gk; 42.55.Px
1 Introduction
Ammonia is an important gas needed in various in-dustrial processes and products, ranging from organic chem-istry and chemical synthesis to fertilizers, refrigerators, chem-ical fiber production and NOx removal in combustion pro-cesses. The widespread use and the toxic nature of ammoniaresults in the need for an adequate gas sensor, which can beused in process control applications as well as for the pre-vention of pollution risks and potential health hazards. Diodelasers are already used as ammonia sensors, for example inindustrial SNCR and SCR DeNOx processes [1]. One mainadvantage of diode-laser gas sensors is the ability to meas-ure reactants and intermediates in situ and non-intrusivelydirectly in the reaction chamber without the need for samplingand recalibration and with high time resolution. Up to now,vertical-cavity surface-emitting lasers (VCSELs) have beenused for absorption spectroscopy of O2 near 760 nm [2–7]
and for H2O measurements near 962 nm [8]. Reference [7] re-ported unsatisfactory results using VCSELs as compared toDFB lasers. None of these papers report highly resolved low-pressure spectra, and Refs. [4] and [8] state that their VCSELsystems were unsuitable for performing low-pressure meas-urements. Only recently have long-wavelength VCSELs beendeveloped [9–13]. The results presented in this paper showthat VCSELs are very well suited not only to measure spec-tra at high pressures, but also for high-resolution spectroscopyat low pressures. A 1.5-µm VCSEL is used to measure highlyresolved spectra of NH3 at various pressures ranging from9.6 mbar to 1 bar. For the first time the high-speed tuning ca-pabilities of near-IR VCSELs are investigated and used forhigh-speed (within 2 µs) gas sensing.
More information about the NH3 absorption lines inthe 1496–1582 nm range, which are best suited for diode-laser combustion–emission and air-quality monitoring, can befound in Webber et al. [14]. The 1.5-µm wavelength regionis also very well suited for the measurement of a variety ofother species ranging from OH, N2O, HCN, HI, H2S to COand CO2 [1, 15].
2 Absorption theory
The transmission T of monochromatic light withfrequency ν (cm−1) though a homogeneous gas of length Lis described by the Beer–Lambert relation T(ν) = I/I0(ν) =exp(−α(ν)L), where Iand I0 are the transmitted and the in-cident spectral intensities, α(ν) is the spectral absorption co-efficient and A = αL = − ln(I/I0) is called the absorbance.The frequency ν in wave number (cm−1) units is related tothe wavelength λ (given in cm) by ν = 1/λ. The spectral ab-sorbance for a gas containing N different species can be cal-culated as sum over all individual transitions i of each speciesby:
A(ν) = PLN∑
g=1
Xg
∑
i
Sg,iϕg,i(ν) (1)
where P (atm) is the total pressure, Xg is the mole fractionof species g, Sg,i (cm−2/atm) is the transition line strengthof transition i of the species g and ϕg,i (cm) is the normal-ized line shape function of this transition (
∫ϕg,i(ν)dν = 1).
Each Sg,i is a function of temperature and each ϕg,i is
604 Applied Physics B – Lasers and Optics
a function of temperature, pressure and gas composition.For low pressures (typically < 10 mbar) the line shapecan be approximated by the Doppler line shape, ϕD(ν) =2/∆νD(ln(2)/π)1/2 exp(−4 ln(2)[(ν− ν0)/∆νD]2), where ν0
is the line-center frequency and ∆νD is the Doppler full-width half-maximum (FWHM) linewidth given by ∆νD =7.1623 ×10−7ν0(T/M)1/2. T (K) is the absolute tempera-ture and M (atomic mass units) is the molecular weight.This line broadening is a consequence of the thermal mo-tion of the gas molecules and of the associated Dopplershifts. For ambient pressure or higher, the linewidth of eachindividual line is dominantly broadened by collisions withother gas molecules. The line shape can be approximated bythe collision-broadened line shape, ϕC = 1/(2π)∆νC/[(ν −ν0)
2 + (∆νC/2)2], where the collisional FWHM linewidth∆νC is given as sum over the different gases in the sampleby ∆νC = P
∑g(Xg2γg) and where γg (cm−1atm−1) are the
collision-broadening coefficients of the different gases withthe gas to be measured. For the pressure range in between theabove two cases, the Voigt line-shape function is commonlyused. The Voigt line shape ϕV is defined as the convolutionof the Doppler and the collisional-broadened line shapes byϕV (ν) = ∫
ϕD(u)ϕC(ν−u)du. See [13] for more details.
3 Description of the device
The measurements were carried out using a novel1.54-µm InGaAlAs/InP VCSEL, which incorporated a buriedtunnel junction (BTJ) as well as a dielectric layer stack whichserved as a back mirror (described in detail in [10]). TheBTJ technology provided the device with effective currentconfinement and waveguiding, which are otherwise difficultto realize in InP-based material systems like (InGaAl)(As)or (InGa)(AsP) that are required for wavelengths around1.5 µm [16]. The current confinement ensured that the re-quired current density in the active region could be obtainedwith an operating current as small as possible and thereforeavoided excess heating. The strong waveguiding associatedwith the BTJ allowed the application of small BTJ diam-eters and therefore stable single-frequency emission withside-mode suppression ratios well exceeding 30 dB. The po-larization of the laser could be controlled using an ellipticallyshaped BTJ to lift the polarization degeneracy. Moreover, theBTJ and the dielectric back mirror allowed a significantly im-proved device design with respect to reduced resistivity andheat generation in the conducting layers and optimized thethermal conductivity between the active region and the heatsink, which is necessary to efficiently operate the active re-gion to obtain a low threshold current as well as high outputpower. 1.55-µm VCSELs fabricated in the buried-tunnel-junction technology have shown sub-mA threshold currents,10–100 Ω series resistance, output powers up to 7 mW (20 C,CW), differential efficiencies > 25% and CW operation upto > 110 C [9]. The large current tuning range for BTJ-VCSELs is a result of the high current density in the activeregion and is not a consequence of the resistive heating in theepitaxial mirror. The VCSEL showed a mode-hop-free con-tinuous single-frequency tuning over the whole laser drivingcurrent range. The concept of InP-based VCSELs using a BTJand a dielectric mirror holds for a wavelength range from
1.3 µm to at least 2 µm [11]. Customizations concerning thedifferent wavelengths determine the appropriate choice of di-electrics, recalculated layer thicknesses and appropriate alloycompositions in the active region.
4 Experimental set-up
The 1.54-µm VCSEL, supplied in a TO-46 case,was mounted on a copper heat sink with integrated thermis-tor and a small Peltier cooler for temperature control. Thelaser emission was collimated using an anti-reflection-coatedaspheric lens with 4.5 mm focal length and < 0.25% nomi-nal reflectivity at 1.5 µm. The temperature controller and laserdiode driver were the HTC-3000 from Wavelength Electron-ics and the LDX-3220 from ILX Lightwave, respectively. The0–7 mA current needed to drive the VCSEL was only a frac-tion of the laser driver output current range of 0–200 mA,but despite this fact the laser driver was of sufficiently lownoise to record clean spectra. The laser beam was directedwith mirrors two times through the 43.7-cm-long samplingcell (total absorption path length 87.4 cm) and onto the DC-10 MHz InGaAs photodetector (PDA400-EC, Thorlabs Inc.)with 1 mm active diameter and a 800–1750 nm response (seeFig. 1). The quartz glass measurement cell, which had 25-mmdiameter wedged fused silica windows, could be heated to1300 K using electric heating shells and could be evacuatedusing a roughing pump in series with a cold trap to 2 mbar.The pressure was measured using a Varian VCMT 13T ca-pacitive pressure transducer and the data shown in the figureswas recorded at 31.25 MS/s sampling rate with a 12-bit Sin-gatec PDA12A data acquisition board mounted in a personalcomputer. Only for the high-speed measurements (Fig. 8, topwindow of Fig. 9) was a different sampling rate of 125 MS/sused. A 5-mm-thick sapphire window (etalon) was positionedin the beam path before each series of experiments. The sap-phire window generated an interference signal (etalon trace)which was used to convert the timescale of the measurementinto a linearized wavelength scale (see Figs. 3 and 8). The NH3
line list by Lundsberg-Nielsen et al. [17, 18] was then usedto shift this relative wavelength scale to the correct absolutewavelength/wave number.
5 Results and discussion
No data averaging or smoothing were appliedfor the figures presented in this paper. The spectra shown
FIGURE 1 Experimental schematic of the 1.5-µm VCSEL absorptionspectroscopy set-up. DAQ, data acquisition board
TOTSCHNIG et al. High-speed VCSEL absorption spectroscopy of ammonia (NH3) near 1.54 µm 605
are single-scan raw data. In Fig. 2, the diode laser volt-age versus injection current and the output power versusinjection current characteristics are shown. At 20 C lasertemperature, the laser threshold was 0.93 mA, the peak out-put power was 0.58 mW (at 6.1 mA) and the series resis-tance was 60 Ω. The series resistance is very low comparedpublished data of other VCSELs [2]. This is a result ofthe buried-tunnel-junction (BTJ) VCSEL technology [16].The low series resistance and the current confinement re-duce unwanted device heating and therefore allow the ef-ficient operation of the active region and a low thresholdcurrent, as well as high output power, to be obtained, evenfor long-wavelength VCSELs. The current tuning rate wasmeasured at 0.7 nm/mA (−2.96 cm−1/mA) and the tem-perature tuning rate was 0.11 nm/K (−0.46 cm−1/K). Inthe top window of Fig. 3, the detector signal is shown forthe laser beam passing twice through the 43.7-cm-longgas cell filled with 9.6 mbar of pure NH3 at room tem-perature. The laser injection current was modulated bya 1-kHz 0–5.8 mA triangular ramp and the laser tempera-ture was 15.2 C. Also shown in the top window is thebaseline used and in the bottom window the etalon in-terference trace (in units of absorbance) used to measure
FIGURE 2 The laser-diode voltage and output power versus laser injectioncurrent characteristics measured at 20 C
FIGURE 3 In the top window the detector signal of a laser absorption ex-periment is given for the gas cell filled with 9.6 mbar of pure NH3. Alsoshown in the top window is the baseline, and in the bottom window the etalontrace
the wavelength tuning of the laser during a scan. The fre-quency jitter of the laser emission between different scans(most likely caused by the noise of the laser driver) wasup to 0.01–0.03 cm−1. Therefore, for very high precisionmeasurements three detectors and beam splitters and anetalon with a low free spectral range should be used tosimultaneously measure the laser power, the etalon trans-mission and the absorption signal. Since the purpose ofthis paper was to demonstrate the measurement capabil-ities of the 1.5-µm VCSEL and not to propose correc-tions to the HITRAN2000 database, the different traceswere measured sequentially. In the top window of Fig. 4,a measured absorbance spectrum of 9.6 mbar of pure NH3is shown. The spectrum in the top window of Fig. 4 isa combination of two overlapping spectra, each recordedby tuning the laser with a triangular 0–5.8 mA currentramp, but at two different laser temperatures. The left partof the spectrum was measured at a laser temperature of15.2 C and is actually the evaluation of the data shownin Fig. 3. The right part of the spectrum was recorded ata laser temperature of −3.6 C. The tuning range of thelaser at 15.2 C, 1 kHz and 0–5.8 mA current modulationwas 12.162 cm−1 (2.9 nm). The range covered by the twooverlapping spectra was 20.4 cm−1 (4.8 nm). In the bot-tom window of Fig. 4 a calculated spectrum of pure NH3is shown for 296 K gas temperature and 9.6 mbar pressure.The spectrum was calculated assuming a uniform pressure-broadening coefficient of γS = 0.3 cm−1atm−1 and Voigtline shapes. The line list of Lundsberg-Nielsen et al. [17,18] was used for the calculation, but unfortunately no in-dividual self-broadening coefficients γS are given for thelines listed. Lundsberg-Nielsen [18] suggests a represen-tative value of γS = 0.3 cm−1atm−1, which was used inthe calculation. The γS of individual NH3 lines typicallyvary between 0.1 and 0.6 cm−1atm−1 and because of thisuncertainty in the self-broadening coefficients, the calcu-lated spectrum (bottom window of Fig. 4) is accurate onlyfor the line positions. It can be seen from Fig. 4 that all
FIGURE 4 The absorbance spectrum of 9.6 mbar of pure NH3 at 296 K and87.4 cm absorption path length, measured with the 1.5-µm VCSEL, is shownin the top window. The calculated spectrum using the Lundsberg-NielsenNH3 line list is shown in the bottom window
606 Applied Physics B – Lasers and Optics
the measured strong lines are listed in the Lundsberg-Nielsen line list, but many weaker lines are missing. Theline strengths listed in Lundsberg-Nielsen et al. [17, 18]are on average 23% lower than the line strengths meas-ured from the data shown in the top window of Fig. 4.This relates well to the work by Webber et al. [14], whichstates that the line strengths listed in Lundsberg-Nielsenare on the average too low by 10%–20% compared totheir own measurements. The spectrum of the isolated6510.986 cm−1 line measured in Fig. 4 (indicated by the ar-row) is shown enlarged in Fig. 5. In Fig. 5, the spectrum ofthe 6510.986 cm−1 NH3 line measured with the 1.537-µmVCSEL is compared with an automated Voigt fit carriedout by peak-fitting software. The Voigt fit gives a shape pa-rameter a = (ln(2))1/2∆νC/∆νD of 0.5, a Doppler linewidth∆νD (FWHM) of 0.01997 cm−1 and a collision-broadenedLorentzian linewidth ∆νC (FWHM) of 0.0119 cm−1. Themeasured Doppler linewidth of 0.01997 cm−1 agrees within3% with the theoretical value calculated by ∆νD =7.1623 ×10−7ν0(T/M)1/2. The measured Lorentzian line-width of 0.0119 cm−1 agrees within 6% with the value cal-culated from the measured self-broadening coefficient γS.This demonstrates the ability of the 1.537-µm VCSEL to ac-curately measure highly resolved low-pressure spectra. Theself-broadening coefficient γS of the isolated 6510.986 cm−1
line was measured to be γS = 0.6 cm−1atm−1 by acquiringa series of spectra of pure NH3 in the gas cell with pres-sures varying from 9 to 145 mbar. In order to further illustratethe ability to measure highly resolved low-pressure spec-tra, spectra of NH3 were measured at pressures of 9.6, 67.5,134, 273, 543 mbar and 1 bar. In order to allow better visualinspection, parts of the whole measured spectra are shownenlarged in Fig. 6. It can be seen that the strong pressurebroadening and the close spacing of the different NH3 linesresults in a blending of lines for pressures above 60 mbar.Only at 9.6 mbar is the pressured broadening reduced toa level where the individual lines can be resolved. For thevery strong absorption feature near 6497.36 cm−1, the meas-urement points with absorbance values of more than three
FIGURE 5 Enlarged part of the measured spectrum shown in the top win-dow of Fig. 4 (indicated by an arrow in Fig. 4). The measured data iscompared with a Voigt fit
are rather noisy, since at these high absorbances almost nolight (I/I0 = exp(−3) = 0.05, i.e. the transmission is less than5%) is transmitted to the detector and the detection noise andthe 12-bit data acquisition limit the accuracy. The 12.2-cm−1
(2.9 nm) tuning range of the VCSEL at 15.2 C was used totest if some strong high-temperature NH3 lines, useable forcombustion monitoring, were accessible by the laser. Figure 7shows spectra of pure NH3 measured at 40 mbar and at 296,468, 606, 1020 and 1173 K gas temperature. Only a small partof the acquired spectrum is shown in order to allow bettervisual inspection. It can be seen that the peak absorption ofsome features (at 6503.7 cm−1 and at 6504.7 cm−1) stronglydecrease with increasing temperature while other absorptionfeatures (at 6540.0 cm−1 and at 6505.3 cm−1) show almost nodecrease between 296 and 468 K, but decrease strongly fortemperatures above 600 K. In the 12.16-cm−1 tuning range,no absorption feature was found which absorbs strongly fortemperatures above 600 K. One main interest point of this
FIGURE 6 Spectra of pure NH3 measured with the 1.5-µm VCSEL for gaspressures of 9.6, 67.5, 134, 273, 543 mbar and 1 bar (order: bottom to top ofgraph)
FIGURE 7 Spectra of 40 mbar pure NH3 measured with the 1.5-µmVCSEL at gas cell temperatures of 296, 468, 606, 1020 and 1173 K (order:top to bottom of graph)
TOTSCHNIG et al. High-speed VCSEL absorption spectroscopy of ammonia (NH3) near 1.54 µm 607
paper was to investigate the high-speed tuning and meas-urement capabilities of 1.54-µm near-infrared VCSELs. InFig. 8, the tuning range of the VCSEL is shown as a functionof the tuning frequency. In order to measure the laser tuning,a 5-mm-thick sapphire window was placed in the beam pathand the interference signal was detected. The free spectralrange (FSR, the frequency difference between two neigh-boring transmission maxima) of the etalon was 0.574 cm−1
(0.136 nm). In the top window (b) of the inset the detector sig-nal without the etalon (baseline) and the detector signal withthe etalon inserted in the beam path are shown for a 120-kHz,0–5.8-mA triangular current modulation. In the bottom win-dow (c) of the inset, the etalon trace defined as the negativelogarithm of the ratio of etalon and baseline signals is shownin absorbance units. The point A indicates the turning pointof the triangular 0–5.8-mA current modulation. As can beseen from the etalon trace, the laser tuning continues beyondpoint A up to point B. This means the laser still heats up evenafter the current has already started to decrease. The tuningrange is defined in this paper as the wavelength tuning be-tween the laser threshold and the point B. From the data shownin the main window (a) of Fig. 8 it can be seen that the tuningrange only slowly decreases with increasing tuning frequencyfrom 13.5 cm−1 (3.2 nm) at 200 Hz to 4.5 cm−1 (1.1 nm) ata tuning rate of 200 kHz (see also Fig. 9). For the 200-kHztriangular current modulation, the time needed to scan overthe 4.5 cm−1 range was only 2 µs. The tuning range couldbe increased if the 0–5.8-mA current modulation amplitudeis increased close to the maximum possible current. In orderto further demonstrate the high-speed measurement capabil-ity of the 1.537 µm VCSEL, in Fig. 9 a 4.5-cm−1-wide NH3spectrum was measured in 2 µs (200-kHz triangular currentmodulation). The measured high-speed spectrum (top win-dow) is compared with a spectrum measured at 1 kHz (bottomwindow). The frequency scale is valid for both windows.The timescale applies only to the top window displaying thehigh-speed measurement. The laser tunes in Fig. 9 from theright to the left. Comparing the timescale and the 1-kHz spec-trum (bottom window), one can see that the laser scans sofast that it tunes over the 6505.262 cm−1 line in less than
FIGURE 8 In window a, the tuning range versus tuning frequency is givenfor a 0–5.8-mA triangular current modulation. Windows b and c explain thetuning measurement
FIGURE 9 In the top window, a 4.5-cm−1-wide spectrum of 40 mbar pureNH3, measured with the 1.5-µm VCSEL in 2 µs, is displayed. In the bot-tom window the same gas sample is measured with the 1.5-µm VCSEL at1-kHz tuning frequency. The frequency scale applies to both windows, thetimescale applies only to the top window. The 10-MHz bandwidth limitationof the detector results in a smoothed high-speed spectrum
50 ns. The peak tuning speed in the middle of the scan wasabout 3.5 cm−1/µs. The 10-MHz bandwidth detector can-not follow the sharp absorption spikes at this speed and asa consequence the high-speed spectrum is broadened by theinsufficient bandwidth detector. At the beginning of the tun-ing and at the end of the tuning, when the laser tunes moreslowly, the detector smoothing is reduced and the measuredpeaks sharpen. Using a higher bandwidth detector, the accu-rate measurement of a 4.5 cm−1-wide spectrum in 2 µs shouldbe possible. Very high time resolution and measurement ratesare therefore possible with near-infrared VCSEL absorptionspectroscopy.
6 Conclusions
For the first time the high-speed tuning capabil-ities of a 1.5-µm VCSEL have been investigated and usedfor high-speed gas sensing. The measured peak tuning speedwas 3.5 cm−1/µs and a 4.5-cm−1-wide NH3 spectrum wasrecorded in 2 µs. At 200-Hz triangular current modulation,the tuning range was 13.5 cm−1 (3.2 nm). A 1.5-µm VCSELwas used to measure highly resolved spectra of NH3 at var-ious pressures ranging from 9.6 mbar to 1 bar. The meas-ured Doppler linewidth of 0.02 cm−1 for the 6510.986 cm−1
line at 9.6 mbar agreed within 3% with the theoretical value.A 20.4-cm−1 (4.8 nm)-wide NH3 spectrum was recorded andcompared with the NH3 line list of Lundsberg-Nielsen. It wasfound, that, in contrary to indications in the referenced liter-ature, VCSELs are very well suited for high-resolution spec-troscopy even at low pressures.
The main advantages of 1.54-µm near-IR VCSELs for in-dustrial gas sensing applications are the large tuning range bylaser injection current (> 10 cm−1 as compared to 1–2 cm−1
for typical distributed-feedback (DFB) diode lasers) and thevery high tuning rates (> 3 cm−1/µs). The broad tuning rangefacilitates multi-species measurements, temperature meas-urements (by measuring a high- and a low-temperature linein one laser scan) and high-pressure measurements (up to
608 Applied Physics B – Lasers and Optics
more than 100 bar). Further advantages of VCSELs as com-pared to edge-emitting diode lasers are their wafer testingcapability, low manufacturing costs, low current operationand therefore low power consumption, less susceptibilityto optical feedback (no optical isolator needed) and lowbeam divergence, which eases fiber coupling. The availabilityof long-wavelength (1.3–2 µm) single-frequency VCSELsshould significantly expand the application fields for infrared-diode-laser gas sensors and make VCSELs a preferred choicefor industrial diode-laser gas-sensing applications.
![Coordination of 9-methyladenine to [cis-Pt(NH3)2]2+ and [Pt(dien)]2+ as studied by proton NMR](https://static.documents.pub/doc/80x56/63250b4f545c645c7f097e50/coordination-of-9-methyladenine-to-cis-ptnh322-and-ptdien2-as-studied.jpg)
![EPR and NMR Spectra as Probes of Spin-Density Distributions in Heterocyclic Ligands Coordinated in trans-[L(Im)(NH3)4RuIII]: Crystal Structure of trans-[(Im)2(NH3)4Ru]Cl3.cntdot.H2O](https://static.documents.pub/doc/80x56/63509fd3275cc44f4c093492/epr-and-nmr-spectra-as-probes-of-spin-density-distributions-in-heterocyclic-ligands.jpg)
![Crystal structure, thermal, electric and magnetic study of [(CH2)7(NH3)2]CoCl2Br2](https://static.documents.pub/doc/80x56/6322f5e0078ed8e56c0aa411/crystal-structure-thermal-electric-and-magnetic-study-of-ch27nh32cocl2br2.jpg)
