NMI TR 3 Characterisation and Calibration of Wavelength Measuring Instruments Based On

NMI TR 3

Characterisation and Calibration of Wavelength Measuring Instruments Based on Grating Spectrometers

Dr Philip B. Lukins

First edition — June 2005

Bradfield Road, Lindfield, NSW 2070

PO Box 264, Lindfield, NSW 2070

Telephone:(61 2) 8467 3600

Facsimile:(61 2) 8467 3610

Web page:

© Commonwealth of Australia 2005

CONTENTS

Preface

1Wavemeter

1.1Reference Laser Sources

1.2Wide-range Calibration

1.3Medium-range Calibration

1.4Narrow-range Calibration

1.5Short-term and Long-term Drift

1.6Conclusions

2Optical Spectrum Analyser

2.1Calibration of the OSA

2.2Calibration of a Multimode Diode Laser – An Example

2.3Conclusions

Acknowledgements

References

Preface

A new laser measurement service is currently being established at NMI. This service will provide measurement of various laser parameters including power, energy, wavelength, linewidth, coherence length, pulse parameters, beam divergence and spatial mode. Customers using this service may require wavelength calibration of their lasers or calibration of their power meters at precisely defined wavelengths. Therefore, appropriate wavelength measuring devices, such as wavemeters and optical spectrum analysers must be used.

This report describes the characterisation and calibration of both a wavemeter and an optical spectrum analyser, and caters for the need for a traceable calibration of wavelength for this calibration service. Both the wavemeter and the optical spectrum analyser are based on grating spectrometers. As we will see in clause1, some interesting and important issues arise in relation to the characterisation and calibration of, in particular, grating-type wavemeters.

1

1Wavemeter

Laser wavemeters are used to measure the wavelength of continuous or pulsed visible or infrared lasers, and are used widely in laser-based metrology and in high-resolution applications in industry and research. Accurate calibration at the 0.001 – 0.01 nm or <10 ppm level is necessary for most applications.

Traditionally, laser wavemeters have been of the interferometric type (eg. Michelson, Fizeau, Mach-Zender). For this type, calibration at a single wavelength is straightforward and generally sufficient. However,wavemeters based on high-order grating spectrometers are now becoming commonplace because of their simplicity, robustness and lower cost.

Both types of wavemeter can suffer systematic errors at the ppm level, and while these are well-known and well-characterised for interferometric-type wavemeters, the same is not true for grating-type wavemeters. Unfortunately, wavelength-dependent non-idealities in the spectrometer and the use of proprietry firmware and optoelectronic designs that are not transparent in terms of their operation means that accurate evaluation and calibration of this type of wavemeter is usually not available and would, in general, require calibration at many wavelengths across the range over which the wavemeter would be used.

As a demonstration of these issues, a commercial grating-type wavemeter (Coherent Inc ‘Wavemaster’ laser wavemeter, serial number WO223 [1]) was calibrated at ~20 laser wavelengths across the range 399 – 935 nm. The wavemeter is based on a grating spectrometer operating in a high-order diffraction mode. A laser beam is coupled to the spectrometer via a 50 m core diameter multimode fibre cable terminated at the instrument end by an ST connector. The input to the fibre is achieved by a post-mounted probe head which has a choice of two angular aperture settings and a rotatable fitting housing a 45 flat silica optic which acts as a beam sampler with a reflectance ~5%. This rotatable beam sampler allows the laser to be coupled into the fibre at either 0 or 90 relative to the direction of the laser beam. The spectrometer incorporates a linear diode array as the light-sensing element, a microcontroller-based acquisition and display system, and an internal He–Ne laser for autocalibration. Firmware is included to calculate the wavelength from the diode array data, calculate related quantities such as frequency, and to apply corrections to compensate for the refractive index of air and spectrometer nonidealities. The units displayed are air wavelength (nm), vacuum wavelength (nm), wavenumber (cm–1) and frequency (GHz). The manufacturer’s specifications [1] are:

• wavelength range 380 – 1095 nm
• wavelength resolution0.001 nm
• wavelength accuracy0.005 nm
• laser linewidth<5 nm
• optical input power20W – 100 mW

This wavemeter is currently used for laser wavelength calibration, wavelength stabilisation of laser/sphere sources and wavelength measurement of diode, ion, dye and Ti:sapphire lasers.

1.1Reference Laser Sources

A Spectra-Physics 165 argon/krypton mixed gas laser was used to obtain the 487.990, 496.512, 501.716, 514.536 and 520.832 nm argon-ion lines and the 476.243, 482.518, 530.866, 568.189, 647.089 and 676.442 nm krypton-ion lines.The blue He–Cd line at 441.565 nm was provided by a Kimmon IK5651R-G laser. Three He–Ne laserswith lines at 632.817, 543.516333 and 611.970770 nm were used: the 633 nm line being from a standard
2 mW Spectra-Physics He–Ne laser while the 543 nm and 612 nm lines being from custom-made I2-stabilised He–Ne lasers in NMI’sLength Group. Reference wavelength values for these lines were obtained from compilations [2–5] of standard laser wavelengths. The uncertainties in the reference wavelengths were 0.005 pm for the I2-stabilised He–Ne lasers and 0.5 pm for the other laser lines.

Four further reference wavelengths were obtained by tuning Ti:sapphire and diode laser systems to resonances of atomic Yb and Yb+ ions in a magnetically-shielded trap in NMI’sTime and Frequency Group.The three transitions used were [2, 6–8]:

• Yb 4f146s2 1S04f146s6p1P1(398.91142 nm)
• Yb+ 4f146s2S1/24f146s2P1/2(369.5243 nm)
• Yb+ 4f145d2 D3/24f145d6s (3D) 3[3/2]1/2(935.186 nm)

The uncertainties in the wavelengths of these transitions are0.01 pm to0.3 pm [2, 9–17]. A frequency-doubled CW Ti:sapphire ring laser system (Coherent 899–21) pumpedby an argon-ion laser (Coherent I400–20) was used to produce radiation at four wavelengths : 739.0482 and 797.82284 nm (fundamental), and 369.5243 and 398.91142 nm (second harmonic). The 369.5241 nm radiation was not measured directly because this wavelength is outside the usable range of the wavemeter, but the relevant transition was still excited by this wavelength so that the fundamental at 739.048 nm could be used as a wavemeter calibration point. An extended-cavity diode laser system was temperature and grating tuned to the 935.186 nm resonance.

Preliminary measurements indicate that the wavemeter firmware uses a dispersion relation for nair() to calculate vac from air. That is, the wavemeter measures air, then calculates nair for this wavelength then evaluates vac from air=vac / nair().

1.2Wide-range Calibration

The wavemeter was calibrated at 19 wavelengths across the range 399 – 935 nm which is almost the whole of its operating range of 380 – 1095 nm. In most cases, airreference wavelength values were used and so the wavemeter was set to read air. In the case of six high-precision reference wavelengths which are quoted for vacuum conditions, the wavemeter was set to read vac.For the measurements using Yb and Yb+ transitions, the trap is, of course, at vacuum so the transitions are detected in vacuo even though the laser is measured at ambient conditions. This approach means that a separate refractive index dispersion correction is not required: this correction is doneby the wavemeter firmware and so is integral to the overall wavemeter accuracy. Reference wavelengths (ref), wavemeter readings (meas) and differences (ref – meas) are shown in Table 1 and Figure 1.The wavelengths are also defined as vac or airin Table 1.

Reference wavelengths were chosen that are well away from atmospheric molecular absorption lines. This eliminates any possibility that absorption and dispersion effects may change the spectral shape of the transmitted laser beam causing a small shift in the effective centroid and hence the wavemeter reading.

The scatter in the data points in Figure 1 is apparently random and uncorrelated. This means that interpolation approaches will not yield an effective improvement in accuracy, nor any useful estimate of the accuracy over ranges of ~1 nm or more. There is no correlation between the accuracy of the reference source and the deviation of the wavemeter reading from the reference wavelength.

The mean difference between reference and measured wavelengths is <ref – meas> = –0.2 pm (standard deviation  = 2.8 pm).These wide-range calibrations showed that the wavemeter readings deviated from the known reference wavelengths by <5 pm.

Table 1. Wide-range wavelength calibration of the wavemeter

Reference wavelength,ref(nm) / Measured wavelength,meas(nm) / ref – meas(nm)
398.91142* / 398.911 / 0.00042
441.565 / 441.569 / –0.004
476.243 / 476.243 / 0
482.518 / 482.519 / –0.001
487.990 / 487.986 / 0.004
496.512 / 496.508 / 0.004
501.716 / 501.716 / 0
514.536 / 514.530 / 0.006
520.832 / 520.836 / –0.004
530.866 / 530.868 / –0.002
543.516333* / 543.515 / 0.0015
568.189 / 568.190 / –0.001
611.970770* / 611.970 / 0.0005
632.817 / 632.817 / 0
647.089 / 647.092 / –0.003
676.442 / 676.444 / –0.002
739.0482* / 739.050 / –0.0018
797.82284* / 797.822 / 0.00084
935.186* / 935.190 / –0.004

*vac values; other wavelengths are air values.

Figure 1. Difference between the reference and wavemeter-measured wavelength
as a function of the reference wavelength

1.3Medium-range Calibration

To evaluate the wavemeter over smaller wavelength ranges of ~1 nm, the frequency-doubled CW Ti:sapphire ring laser system was again used and tuned to the 399 nm Yb transition as in clause 1.2. In this calibration, the laser was then tuned away from the transition by a single longitudinal mode (thin etalon mode spacing of 30 GHz) at a time and the fundamental laser wavelength near 798 nm measured using the wavemeter. The results are shown in Table 2 and Figure 2.

Since the laser mode spacing, and hence the laser wavelength shift, is fixed, the incremental wavelength difference as given by the wavemeter is a measure of the ability of the wavemeter to measure small sequential wavelength differences accurately. The mean incremental difference is 21.0 pm (standard deviation  = 0.8 pm). There is a single outlying data point at 24 pm due to residual wavemeter systematic uncertainties. These medium-range calibrations indicated wavemeter deviations of typically2 pm over a range of ~1 nm.

1.4Narrow-range Calibration

The 399 nm transition in Yb neutrals has a width of ~3 GHz.This transition has six hyperfine lines due to 171Yb and 173Yb with a splitting of <1 GHz [6]. The laser was tuned to each of these hyperfine transitions using the fluorescence visible in the trap as an indicator. The six hyperfine transitions occur at 398.910 – 398.911 nm as measured by the wavemeter corresponding to a fundamental laser beam at 797.820 – 797.822 nm. These narrow-range measurements indicated wavemeter deviations of <1 pm over wavelength ranges of a few pm.

Figure 2. Wavemeter readings as a function of the actual wavelength expressed as consecutive longitudinal laser modes with spacings of 30 GHz

Table 2. Medium-range calibration of the wavemeter

Laser mode number n
(nm)* / Measured wavelength
(nm) / Incremental difference
(nm)#
1 / 797.339 / –
2 / 797.359 / 0.020
3 / 797.380 / 0.021
4 / 797.401 / 0.021
5 / 797.425 / 0.024
6 / 797.446 / 0.021
7 / 797.467 / 0.021
8 / 797.488 / 0.021
9 / 797.509 / 0.021
10 / 797.530 / 0.021
11 / 797.550 / 0.020
12 / 797.571 / 0.021
13 / 797.593 / 0.022
14 / 797.614 / 0.021
15 / 797.635 / 0.021
16 / 797.655 / 0.020
17 / 797.676 / 0.021
18 / 797.697 / 0.021
19 / 797.718 / 0.021
20 / 797.740 / 0.022
21 / 797.760 / 0.020
22 / 797.782 / 0.022
23 / 797.802 / 0.020
24 / 797.823 / 0.021
25 / 797.844 / 0.021
26 / 797.865 / 0.021
27 / 797.887 / 0.022
28 / 797.907 / 0.020
29 / 797.928 / 0.021
30 / 797.949 / 0.021
31 / 797.970 / 0.021
32 / 797.991 / 0.021
33 / 798.012 / 0.021

*Laser mode number (thin etalon) relative to the shortest wavelength used; Yb resonance occurs at n = 24.

#(n+1–n) where is the wavelength measured by the wavemeter and n is the laser mode number.

1.5Short-term and Long-term Drift

The short-term (10 min) drift and reproducibility of the wavemeter was 1 pm for all of the wavelength measurements.

Possible variations in the calibration of the wavemeter were checked by periodically recalibrating over a period of 18 months at 476.243, 487.990, 514.536, 530.866, 632.817, 647.089 and 676.442 nm. Not surprisingly, the calibration at 632.817 nm was both stable and accurate to 1 pm partly because the wavemeter uses an internal He–Ne laser for self-calibration. For the other wavelengths at which periodic recalibration was performed, the wavemeter readings also changed by no more than 1 pm over 18 months. These long-term variations were random, uncorrelated and near the resolution limit. No systematic trend was seen.

1.6Conclusions

This wavemeter is accurate to 5 pm over its full operating range of 380 – 1095 nm with no apparent correlation between uncertainties for wavelengths separated by more than a few nm. This result confirms the manufacturer’s specification for accuracy [1]. Interestingly, however, its reproducibility for wavelengths within a band of ~1 nm is typically within 1 pm. Furthermore, if the wavemeter is calibrated against a reference laser line then the apparent uncertainty in subsequent wavelength measurements within ~1 nm of this reference calibration is <2 pm. This demonstrates that enhanced accuracy can be obtained by calibrating the wavemeter against reference laser lines within ~1 nm of the wavelength with which the wavemeter would subsequently be used. The origin of this enhanced narrow-range and medium-range performance probably lies in the wavemeter’s optoelectronic design or firmware effectively integrating fluctuations over this ~1 nm range.

The long-term drift results suggest that the wavemeter is reproducible to ~2 pm over several years. For a given wavelength, we therefore recommend a recalibration interval of three to five years. However, grating-type wavemeters should be calibrated for the specific wavelengths for which they are to be used or should be calibrated at several wavelengths across their usable range if they are to be used for broadband measurements.

2Optical Spectrum Analyser

While many lasers are single-line sources, there are an increasing number of broadband (eg. dye, crystalline and diode lasers, and ASE sources) and multiline (eg. diode lasers and some crystalline lasers such as Nd:YVO4) sources. These lasers cannot be measured directly with a wavemeter because of their multiline character which would lead to a potentially erroneous result. This means that an optical spectrum analyser (OSA) is required to measure such sources. OSAs display the emission spectrum of the source rather than a numerical value of the wavelength of the laser line as in the case of a wavemeter. Typically, wavemeters have accuracies ~1 pm while OSAs have accuracies ~10 – 100 pm. Despite their lesser accuracy, OSAs display the actual laser emission spectrum. OSAs can be calibrated against either wavemeters or reference laser lines. Here we perform a calibration of an Anritsu MS96A OSA (serial no. M22691); which has a resolution of 0.1 nm and a sensitivity of ~100 pW over the range 0.6 – 1.6 m.

2.1Calibration of the OSA

The OSA was calibrated against known laser lines from a 5 mW 1523.0 nm Melles–Griot 05-LIP-171 He–Ne laser and a 100 mW 1064 nm SUWTech DPIR-3100 Nd:YVO4 laser. The 1523.0 nm He–Ne line gave a reading of 1522.95 nm while the 1064 nm laser doublet in Nd:YVO4 gave readings of 1064.205 nm and 1064.36 nm. The deviations of the OSA results from the known laser wavelengths were in the range 0.05 to 0.1 nm.

2.2Calibration of a Multimode Diode Laser —An Example

We evaluated a 2 mW Fujitsu fibre-coupled single-emitter multiple longitudinal mode diode laser operating near 1.3 m. It was necessary to measure the laser spectrum to establish an effective wavelength for InGaAs detector calibrations [18]. The laser spectrum consisted of 13 detectable modes and the wavelengths and peak powers for each of these modes were measured. An initial estimate of the effective wavelength was obtained by calculating the power-weighted average wavelength which was 1294.8 nm. The peak powers were then plotted against wavelength and fitted. The mode distribution was symmetric and gaussian. The fitted parameters at 21.5C are:

• centre wavelength1294.91 0.12 nm
• envelope width1.950.11 nm
• mode spacing0.772  0.022 nm
• mode line width0.04  0.01 nm

2.3Conclusions

There is agreement between the known laser lines and the OSA to within 0.1 nm which agrees with the OSA manufacturer’s specification.At this uncertainty level, grating spectrometer based systems are generally linear so that this uncertainty of 0.1 nm is very likely to apply across the wavelength range of the OSA of 600 – 1600 nm.

Acknowledgements

The author is indebted to Dr Bruce Warrington for valuable discussions and the use of the Ti:sapphire and external cavity diode lasers, and to Dr Nick Brown for the use of the I2-stabilised He–Ne lasers.

References

[1]Coherent WavemasterTMUsers Manual, Coherent Inc. Auburn Division (2000)

[2]W.C. Martin, R. Zalubas and L. Hagan, Atomic Energy Levels — The Rare Earth Elements, NSRDS-NBS 60, US Government Printing Office(1978)

[3]CRC Handbook, Vol. 2

[4]E. Jaatinen and N. Brown, Metrologia32, 95 (1995)

[5]T.J. Quinn, Metrologia36, 211 (1999)

[6]A. Banerjee et al., Europhys. Lett. 63, 340 (2003)

[7]A.M. Martensson-Pendrill et al. Phys. Rev. A49, 3351 (1994)

[8]M.J. Sellars et al., unpublished data

[9]V. Kaufman and J. Sugar, J. Opt. Soc. Am.63, 1168 (1973)

[10]J.-F. Wyart and P. Camus, Physica Scripta20, 43 (1979)

[11]P. Taylor et al., Phys. Rev. A56, 2699 (1997)

[12]P. Taylor et al., Phys. Rev. A60, 2829 (1999)

[13]M. Roberts et al., Phys. Rev. A60, 2867 (1999)

[14]M. Roberts et al., Phys. Rev. A62, 020501 (2000)

[15]J. Stenger et al., Opt. Lett.26, 1589 (2001)

[16]S.A. Webster et al., Phys. Rev. A65, 052501 (2002)

[17]P.J. Blythe et al., Phys. Rev. A67, 020501 (2003)

[18]P.B. Lukins, CSIRO Technical Report TIPP-1699 (2003)

NMI TR 31