Wavelength references for 1300-nm wavelength-division multiplexing

776 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 5, MAY 2002

Wavelength References for 1300-nm

Wavelength-Division Multiplexing

T. Dennis, Member, OSA, E. A. Curtis, C. W. Oates, Member, OSA,

L. Hollberg, Associate Member, IEEE, Member, OSA, and S. L. Gilbert, Member, OSA

Abstract—We have conducted a study of potential wavelength

calibration references for use as both moderate-accuracy transfer

standards and high-accuracy National Institute of Standards and

Technology (NIST) internal references in the 1280–1320-nm wave-

length-division-multiplexing region. We found that most atomic

and molecular absorption lines in this region are not ideal for use

as wavelength references owing to factors such as weak absorp-

tion, complex spectra, or special requirements (for example, fre-

quency-doubling orexcitation with an additional light or discharge

source). We have demonstrated one of the simpler schemes con-

sisting of a tunable diode laser stabilized to a Doppler-broadened

methane absorption line. By conducting a beat-note comparison of

this reference to a calcium-based optical frequency standard, we

measured the methane line center with an expanded uncertainty

)of 2.3 MHz. This methane-stabilized laser now serves as a

NIST internal reference.

Index Terms—Absorbing media, optical fiber communica-

tion, optical materials, optical propagation in absorbing media,

optical spectroscopy, semiconductor lasers, standards, wave-

length-division multiplexing.

I. INTRODUCTION

AVELENGTH-division multiplexing (WDM) is rapidly

expanding the capacity of optical fiber communications

systems. Current systems operate in the 1540–1560-nm region,

but WDM will likely expand into other wavelength regions

as well, possibly covering the entire range from about 1280

to 1630 nm. Wavelength calibration references are needed to

calibrate instrumentation and reliably separate densely spaced

channels in these new regions. Following the development of

wavelength calibration references in the 1510–1560-nm region

[1]–[3] and the WDM L-band (approximately 1565–1625

nm) [4], we have begun to investigate potential wavelength

references in the 1300-nm region.

Our goals are to produce high-accuracy references for Na-

tional Institute of Standards and Technology (NIST) internal

calibration and moderate-accuracy transfer standards (such as

NIST Standard Reference Materials) to help industry calibrate

instrumentation. A wavelength reference for calibration in a

standards laboratory should have a frequency uncertainty of less

than 10 MHz, whereas a transfer standard can have an uncer-

tainty of a few hundred megahertz. Atomic and molecular ab-

Manuscript received October 23, 2001; revised February 26, 2002.

T. Dennis, C. W. Oates, L. Hollberg, and S. L. Gilbert are with the National

Institute of Standards and Technology, Boulder, CO 80305 USA.

E. A. Curtis is with the National Institute of Standards and Technology,

Boulder, CO 80305 USA and also with the Physics Department, University of

Colorado at Boulder, Boulder, CO 80309 USA.

Publisher Item Identifier S 0733-8724(02)05119-8.

sorption lines provide wavelength references that are very stable

under changing environmental conditions and have well under-

stood physical behavior. A variety of molecules have distinctive

absorption features in the 1300-nm region due to their quan-

tized vibrational and rotational motion. These transitions are

combination or overtone bands that can be probed directly, but

they typically have low absorption strengths. Atomic transitions

in this region occur between excited states and, thus, require

initial excitation by a laser or electric discharge. Other atomic

or molecular references can be realized by frequency doubling

1300-nm light to probe atomic transitions in the 650-nm region.

Depending on the physical conditions of the reference and the

measurement technique, a wide range of absorption linewidths

can be realized. For example, a molecular or atomic gas at a

pressure of 1 kPa (8 torr) typically has transition linewidths of

less than 1 GHz that are dominated by Doppler broadening. On

the other hand, a gas at a pressure of 100 kPa (

1 atmosphere)

can have pressure-broadened linewidths of more than 10 GHz.

Reductions in the linewidth below the Doppler limit, to widths

of 10 MHz or less, can be realized with saturated absorption

spectroscopy. For weak molecular transitions, this technique

often requires a resonant cavity to increase the laser power.

We have conducted a survey of potential wavelength cal-

ibration references for use as moderate-accuracy transfer

standards and high-accuracy NIST internal references in the

1280–1320-nm WDM region. We then produced a high-accu-

racy reference using one of the simpler schemes consisting of a

tunable diode laser stabilized to a Doppler-broadened methane

absorption line. In Section II, we present the results of our

survey of potential references. In Section III, we describe the

methane-based wavelength reference that we constructed and

compared to a calcium-based optical frequency standard.

II. R

EFERENCES IN THE 1300-nm REGION

Table I summarizes the atomic and molecular species we have

considered as wavelength references for the 1280–1320-nm re-

gion. This range includes and is centered upon the gain spec-

trum of currently available praseodymium-doped fiber ampli-

fiers. Some materials have been listed with specific transitions

and wavelengths, while others are listed with band designations

and wavelength ranges owing to the complexity of the molec-

ular spectra. In some cases, weaker transitions and/or ones of

lower accuracy have been omitted in spectrally dense regions.

The uncertainty of the specified transitions or bands are given as

fractions, where 1

10 at 1300 nm is equivalent to 230 MHz,

1.3 pm, and 8

10 cm . The column of special notes lists

DENNIS et al.: WAVELENGTH REFERENCES FOR 1300-nm WDM 777

TABLE I

UMMARY OF MATERIALS FOR WAVELENGTH REFERENCES AT 1300 nm

some of the principle experimental difficulties that would be en-

countered when working with a material.

The rubidium atom has been used extensively in high-res-

olution spectroscopy. The D2 transition at 780 nm forms a

convenient 1560-nm reference through frequency-doubling

[5], and serves as a NIST high-accuracy reference for this

WDM region [1], [2]. A possibility for a 1300-nm reference

is the 5P

6S transition occurring at 1323.87 nm

between excited states [6]. This transition is slightly outside

the desired wavelength region, and stabilized optical pumping

of the 5P

state is required. However, in one experiment

a distributed-feedback diode laser has been locked to this

transition [7].

The noble gases neon, argon, krypton, and xenon have been

studied extensively as wavelength references in the 1300- and

1550-nm regions [8]–[10]. The materials have benign physical

properties and offer a wide choice of absorption lines, as shown

in the table (all the xenon transitions lie outside our region of

interest). Unfortunately, all the noble gas transitions occur be-

tween excited states,typically requiringa high voltage discharge

of the gas. This rather chaotic excitation process may be accept-

able for moderate-accuracy references, but a detailed character-

ization of the discharge parameters and wavelengthshifts would

likely be required for a high-accuracy system.

Despite its simple chemical structure, the iodine molecule

has an extremely complicated spectrum with more than 20 000

visible absorption lines in the X

B band. A variety of

gas lasers in the visible spectrum have been stabilized to

these transitions, most notably the helium–neon laser at 633

nm. Frequency doubling would allow a 1300-nm laser to be

referenced to one of these lines. A select number of appropriate

lines have been measured to very high accuracy, with hyperfine

components identified [11]. Doppler-limited locking to iodine

has been performed using periodically poled lithium niobate

and an Nd : YAG laser at 1319 nm [12]. In addition to efficient

frequency doubling, this scheme requires moderate heating

to increase the vapor pressure of iodine and populate higher

vibrational levels of the ground state.

The common atmospheric molecules of carbon dioxide and

water have been studied extensively with a combination of

Fourier transform spectroscopy and theoretical calculations

[13], [14]. Most of the absorption lines in the 1300-nm region

are combinations or overtones of fundamental transitions at

longer wavelengths, causing the line strengths to be weak.

As a result, required absorption paths are long, often tens to

hundreds of meters, and resonant cavities are a necessity for

saturated absorption spectroscopy. Compared to water, the

spectrum of carbon dioxide is simple and the lines are easily

778 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 5, MAY 2002

identified, although they are about a factor of 100 weaker

in strength [13]. Absorption lines of water have been used

for spectroscopic calibration [15], [16]; moderate heating to

increase vapor pressure may be necessary.

We also considered the molecules acetylene, ammonia, and

the nitrate radical NO

. Acetylene, which forms a convenient

wavelength reference and is used as a NIST Standard Refer-

ence Material (SRM) between 1513 and 1541 nm [2], [3], has

weak combination transitions near 1300 nm that require absorp-

tion paths of 30 m or more [17]. Even in the 1500-nm region,

where the transitions are stronger, acetylene requires resonant

cavities for saturated absorption spectroscopy [18]. There ap-

pear to be few studies of ammonia at 1300 nm; the long ab-

sorption path that was reported in one investigation is indicative

of weak spectral features [16]. Also, based on spectra collected

between 1450 and 1550 nm, the spectrum at 1300 nm is likely

to be far more complicated than indicated by the four lines that

have been reported. The nitrate radical NO

has been the subject

of numerous studies of chemical structure primarily involving

spectroscopy at visible wavelengths. One study does report on

hundreds of spectral features at wavelengths above 1315 nm; in-

terestingly, water spectra were used for wavelength calibration

in that study [19].

Hydrogen fluoride is a rare example of a molecule that of-

fers strong absorption lines in the 1300-nm region [20]. While

the lines are several orders of magnitude stronger than those of

water, there are only five located between 1280 and 1320 nm.

The primary disadvantage of this material is its high reactivity,

which causes handling difficulties. Most absorption cells for hy-

drogen fluoride have been constructed of metal or plastic with

sapphire or polyethylene windows. Despite this technical dif-

ficulty, diode laser stabilization to the R-branch lines has been

demonstrated [21].

The hydrogen sulfidemolecule offers several hundred lines in

the 1300-nm region, but most of the detailed spectroscopy has

been performed at wavelengths beyond 1600 nm. Therefore, in-

formation at shorter wavelengths on this toxic molecule is lim-

ited. In one relevant study, a Fourier transform spectrometer has

been used to identify the bands near 1300 nm, determine the

band center wavelengths, and count the number of lines within

the bands [22]. The long absorption paths used by the study (28

and 433 m) indicate that the lines are weak. Because individual

line positions have not yet been reported, detailed spectroscopy

would be necessary to form a wavelength reference from this

material.

Methane has been studied extensively with infrared spec-

troscopy to monitor gaseous concentrations in industrial

environments, where it poses a high explosion danger, and in

the atmosphere where it could be important to global warming.

Detection of trace gases requires moderate wavelength accuracy

to identify a species, placing higher emphasis on the accurate

measurement of absorption depth, which is used to determine

gas concentration. For this reason, the wavelength accuracy

of previous studies would need to be improved to form a

high-accuracy reference. The R branch lines of the

band of methane, extending from 1314 to 1329 nm, are well

suited for telecommunications wavelength references [15].

Unfortunately, the region below 1314 nm lacks any significant

absorption features. Nevertheless, methane is a good reference

material because the R branch lines are reasonably strong,

requiring absorption paths of less than 1 m, and can be probed

at room temperature and low pressures. The relative ease with

which the methane spectroscopy can be performed allows for

the construction of an experimental apparatus that is compact,

portable, and robust. Furthermore, methane is easy to handle

in the laboratory, is readily available commercially with high

purity, and can be safely contained in glass cells. Based on

a report of Doppler-free spectroscopy in the 1600-nm region

[23], saturation of the lines at 1300 nm should be possible

using a resonant cavity. In comparison to the clean spectra

of acetylene and hydrogen cyanide used for NIST SRMs in

the 1500-nm region, the spectrum of methane is somewhat

complicated. This is not a serious concern for a high-accuracy

reference because the detailed structure can be resolved at

lower pressures.

Based on the factors discussed in this section, we decided to

produce a high-accuracy wavelength reference using a Doppler-

broadened methane absorption line. This provides a relatively

simple reference that could be measured relative to a highly ac-

curate calcium frequency standard.

III. M

ETHANE HIGH-ACCURACY REFERENCE

Our development of a high-accuracy reference based on a

methaneabsorption line near 1314 nm involvedtwoparts: 1) sta-

bilization of a diode laser to the methane line and 2) the accurate

measurement of the frequency of the stabilized laser. The latter

was necessary because we require a reference with a fractional

uncertainty of less than 1

10 , but the methane lines had

only been reported previously with an uncertainty of 4

10 .

The frequency measurement was accomplished by conducting

a beat-note measurement of the methane-stabilized laser with a

laser referenced to a calcium-based optical frequency standard

developed at NIST.

A. Experimental Setup

Fig. 1 shows the components of the methane-stabilized

wavelength reference we constructed. We used a commercially

available extended-cavity diode laser (ECDL) system having a

100-nm tuning range centered at 1304 nm. The laser linewidth

was measured to be less than 5 MHz in a 50-ms interval with

a Fabry–Pérot spectrum analyzer. The free-space beam of the

tunable laser was passed through a bulk isolator before it was

coupled into the FC/APC connector of a single-mode optical

fiber. The laser power was then passed through a fiber isolator

and divided by a fiber splitter, routing approximately 15% of

the power to the methane stabilization components and 85% to

a beat-note system for the measurement of the laser frequency.

The portion of the laser beam used for methane stabilization

was collimated into a free-space beam and divided by a wedged

beam splitter. One part was directed to a photodiode to monitor

intensity while the other passed through the methane gas cell

to the signal photodiode. Common-mode intensity variations,

such as those due to fluctuations in laser power and to etalon

fringes associated with the diode laser chip and the free-space

isolator, were removed from the methane spectra by subtracting

DENNIS et al.: WAVELENGTH REFERENCES FOR 1300-nm WDM 779

Fig. 1. The 1314-nm methane-stabilized laser and beat-note measurement

system. ECLD: extended cavity laser diode. ISO: isolator. COL: fiber

collimator. BS: beamsplitter. PD1, PD2: photodiodes. PC: polarization

controller. RX: photoreceiver. LO: local oscillator. RF-SA: radio-frequency

spectrum analyzer. G: electrical gain. The dashed signal paths represent

free-space optical beams and the solid paths indicate fiber or electrical

connections.

the photodiode signals. This improved the signal-to-noise ratio

(SNR) and reduced the sensitivity of the stabilization lock-point

to shifts of the fringes caused by temperature fluctuations and

thermal expansion. Also, the windows of the gas cell were

wedged to avoid production of interference fringes. To increase

the absorption, the free-space beam traversed the cell three

times for a total path of 60 cm. The entire optical and electrical

system that formed the methane reference was designed to fit

on a small pushcart to facilitate the beat-note measurements

with the calcium frequency standard.

Fig. 2 shows a transmission spectrum of methane recorded

by an optical spectrum analyzer, with the major R-branch lines

identified. Even at a resolution of 15 pm and a relatively high

gas pressure of 52.0 kPa (390 torr), it is clear that the spectrum

of methane is complicated, with most lines possessing substruc-

ture. Despite the singlet appearance of R(3) in Fig. 2, a detailed

laser scan at 0.3-pm resolution and 6.7 kPa (50 torr) revealed

that it is actually a triplet line with features separated by about

12 pm. By contrast, the R(1) line was found to be a singlet under

the same experimental conditions.

We dithered the center frequency of the laser by

15 MHz

using a 1.6-kHz external modulation signal applied to the laser’s

piezoelectric tuning element. When the laser wavelength was

slowly tuned across an absorption line and the photodiode sig-

nals were monitored with phase-sensitive detection at this mod-

ulation frequency, the derivative of the line shape was mapped.

Fig. 3 shows derivative spectroscopy of the central feature of

the R(8) cluster of lines at a pressure of 1.35 kPa (10.1 torr).

Here, the laser wastuned in

0.3-pmwavelengthstepswhile the

phase-sensitive detection signal was recorded in a 60-Hz band-

width. With the wavelength stepping turned off and the modu-

lated laser tuned near the center of an absorption line, the bipolar

derivative signal represented the offset (error) of the laser from

the zero crossing at line center. This normalized error signal was

conditioned with integral and proportional feedback circuits and

Fig. 2. Spectrum of the R-branch of methane at 52.0 kPa(390 torr) collectedat

a resolution of 15 pm on a vacuum wavelength scale. The spectrum is complex;

most lines shown have substructure when viewed at higher resolution and lower

pressure.

applied to the laser in summation with the external frequency

modulation. Increasing the gain of the feedback circuit forced

the laser to minimize the error signal, locking the wavelength to

the center of the absorption line.

We locked the diode laser to the zero crossing at 1314.588 nm

(vacuumwavelength)in the R(8) cluster of lines shownin Fig. 3.

At low to moderate pressure, this line is sufficiently distinct that

the wings of the two neighboring features at longer wavelengths

have negligible effect on the lock point. We also confirmed with

computermodeling that the small feature at10 pm towardlonger

wavelength causes a negligible shift of the lock point of about

300 kHz. The 1314.588-nm feature was also the cleanest line

in reasonable proximity (

0.3 nm) to the calcium-based optical

frequencystandard, resulting in a beat-note frequency that could

be measured in the microwave range.

To characterize the pressure shift of this methane line, we

stabilized the laser at two gas pressures using interchangeable

cells with low (1.35 kPa; 10.1 torr) and high (8.21 kPa; 61.6

torr) pressures. The absorption depth of the line at the low and

high pressures was 2% and 8%, respectively. We also collected

absorption spectra of the stabilization feature at the two pres-

sures and fitted the results to a Voigt line shape to characterize

the pressure broadening. The low-pressure linewidth was essen-

tially equal to the 720-MHz Doppler-broadened value while the

high-pressure line was broadened slightly to 780 MHz. By con-

trast, the lines of Fig. 2 at an even higher pressure of 52.0 kPa

(390 torr) have widths of about 2.8 GHz.

B. Frequency Measurement

To accurately measure the frequency of the methane-stabi-

lized laser, we compared it to a frequency-doubled 1314-nm

laser that was locked to an optical frequency standard based

on laser-cooled calcium atoms. The calcium frequency stan-

dard has been described in detail elsewhere [24]. In summary,

it consists of an ECDL locked to the narrow (400-Hz natural

linewidth)

S ( ) P ( ) transition at 657

780 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 5, MAY 2002

Fig. 3. Derivative spectroscopy signal of the central feature of the R(8) line

plotted versus vacuum wavelength. The feature with a zero crossingat 1314.588

nm was used to create the methane-stabilized reference.

nm in neutral calcium. In order to achieve subkilohertz spec-

troscopic resolution and greatly reduced systematic shifts, this

standard uses laser-cooled atoms from a magnetooptic trap op-

erating at 423 nm. Recent measurements of the absolute fre-

quency of a laser locked to this transition yielded an uncertainty

of 26 Hz at 657 nm, corresponding to 13 Hz at 1314 nm [25].

Thisobviously far exceedsthe requirements of a wavelengthref-

erence for WDM communications. The second harmonic of an

ECDL at 1314 nm (different from the methane-stabilized laser

and not shown in Fig. 1) was offset phase-locked to the cal-

cium optical standard. By single-passing a periodically poled

LiNbO

crystal with 5 mW of infrared light, more than 300 nW

of second harmonic light at 657 nm was generated for use in the

phase-lock beat signal. The remaining 1314-nm ECDL output

(

3 mW) was then available for generating a beat note with the

methane-stabilized system.

The components of the beat-note measurement system used

for the comparison are shown in the lower half of Fig. 1. The

two sources were combined with a 3-dB optical fiber coupler,

resulting in about 1 mW of optical power from each source. The

fiber from the methane reference contained a fiber polarization

controller to match the two polarization states. Combining the

two sources on the photoreceiver allowed the 225-THz optical

carriers to beat, producing an electrical signal at a differencefre-

quency of about 57 GHz. The photoreceiver had a 3-dB band-

width of 25 GHz, with a measured response at 50 GHz attenu-

ated an additional 9 dB. We sent the output of the photoreceiver

to a harmonic mixer, which mixed the input with harmonics it

generated from the narrow-bandsignal of a synthesized local os-

cillator (LO). The mixing process shifted the center frequency

of the beat note to less than 200 MHz, where it could be readily

measured on a radio-frequency spectrum analyzer.

Fig. 4 shows a typical mixed-down beat-note signal recorded

by the RF spectrum analyzer while the methane-stabilized ref-

erence was locked to the high-pressure methane cell. The signal

was collected with a sweep time of 5 s at a resolution band-

width of 1 MHz. The

30-MHz width of the spectrum was

Fig. 4. Typical heterodyne beat spectrum between the 1314-nm laser locked

to the methane line and the calcium-based frequency reference. The methane

cell pressure was 8.21 kPa (61.6 torr). The mixed-down beat note was viewed

on a RF spectrum analyzer with a 5-s sweep time and a 1-MHz resolution

bandwidth. A center frequency of 163.0 MHz was obtained from this spectrum

with a Gaussian curve fit.

caused by the large frequency dither needed to lock the laser

diode to the Doppler-broadened methane line. In addition to the

dither occurring at a rate of 1.6 kHz, the beat note also had a

slower frequency jitter of about 10 MHz occurring on a time

scale of more than 5 s. We suspect that this was due to a com-

bination of the noise in the laser driver and the limitations of

the feedback locking circuit. Unfortunately, the broad spectral

width of the signal in combination with the moderate SNR of

19 dB (in a detection bandwidth of 200 MHz) made elec-

tronic counting of the center frequency unreliable. Instead, we

obtained a center value of 163.0 MHz for the signal of Fig. 4 by

applying a Gaussian curve fit.

The frequency of the local oscillator driving the harmonic

mixer during the measurement had a value of

MHz, as measured by an electronic counter. The order of the

harmonic that was mixing with the beat note was determined

by dividing the

57-GHz separation of the lasers by . The

order was also determined by observing the difference between

two different local oscillator frequencies that gave the same

center frequency for the mixed-down beat note. From these two

observations, we conclude that the 12th harmonic of

was

mixing with the 57-GHz beat note produced by the photore-

ceiver. Because the center frequency of the mixed-down beat

note increased as

increased, the 12th harmonic of the local

oscillator was situated above the beat frequency. Therefore, the

center frequency of the beat note of Fig. 4 before mixing was

MHz MHz MHz.

The absolute measurement of the frequency of the methane-

stabilized laser was accomplished by accounting for two fre-

quency offsets of the 1314-nm calcium-referenced laser with re-

spect to the calcium line center. The probe laser at 657 nm was

locked to the calcium line with an 80.0-MHz offset while 55.0

MHz separated the doubled 1314-nm light from the probe laser.

Also, the methane absorption line at 1314.588 nm resides at a

higher frequency than the calcium-based reference at 1314.919

DENNIS et al.: WAVELENGTH REFERENCES FOR 1300-nm WDM 781

TABLE II

EASURED FREQUENCIES AND CALCULATED WAVELENGTHS FOR THE METHANE

REFERENCE,

ITH EXPANDED UNCERTAINTIES (2

)

nm. Considering all this, the absolute frequency of the methane

wavelength reference is given by

MHz MHz

where MHz is the calcium transition fre-

quency. The beat-note measurement and analysis were repeated

with the laser stabilized to the low-pressure methane cell.

The stability characteristics of the methane-stabilized laser

were tested in several ways. To demonstrate repeatability,

beat-note measurements were performed again two days later,

during a different part of the laboratory’s daily temperature

cycle (approximately a few degree Celsius). On each day, the

beat note was measured up to ten times to obtain daily averages;

we observed differences in the daily averages of about 1.1 MHz

for both the high- and low-pressure cells. The temperature

sensitivity was investigated in more detail by heating compo-

nents of the system. Temperature changes of up to 5

C did not

cause observable shifts of the beat note frequency. If there was

an effect due to temperature it was negligible relative to the

random uncertainty. We also unlocked and relocked the diode

laser a number of times to demonstrate short-term repeatability,

and we varied parameters such as the amount of dither and

feedback gain. In all cases, the wavelength reference remained

consistent, with fluctuations that were within the uncertainty

of the beat-note measurement.

Table II summarizes the absolute measurement results of

center frequency for the methane reference at 1314.588 nm.

By comparing the high- and low-pressure center frequencies,

we conclude that a 6.87-kPa (51.5 torr) reduction in pressure

resulted in a 49.4-MHz shift of the spectroscopic feature

toward higher frequency. Assuming the pressure shift is linear,

the slope is

7.2 1.1 MHz/kPa ( 0.96 0.14 MHz/torr).

The table also shows the position of the spectroscopic feature

without the pressure shift, as determined by a linear extrapola-

tion to zero pressure. The corresponding center wavelengths in

vacuum were calculated using the defined value for the speed

of light (

m/s).

The uncertainties reported in Table II are the expanded uncer-

tainties with a coverage factor of 2 (i.e., our values are 2

), and

were estimated from a variety of sources. A significant contri-

bution to the frequency uncertainties for the high- and low-pres-

sure cells originates from the standard deviation of the mea-

surements. The center frequency of the high-pressure cell is

the average of 30 measurements, with a standard deviation of

1.0 MHz. The low-pressure value is the result of 16 measure-

ments with a standard deviation of 1.1 MHz. Both of the stan-

dard deviations (of the distributions) account for the daily re-

peatability, the temperature dependent variation, and the various

fitting procedures. The other significant component of the fre-

quency uncertainties arises from the drift in the balance of the

photodetectors and the offset voltages of the methane stabiliza-

tion feedback electronics. We estimate that the zero crossing of

the locking feature at 1314.588 nm was established and main-

tained to within an offset of 5-mV DC. Given the slopes of

the derivative spectroscopy signals at the high and low pres-

sures, the estimated offset corresponds to frequency uncertain-

ties of 0.5 and 1.5 MHz, respectively. The low-pressure mea-

surement has a greater uncertainty because of the weaker ab-

sorption. Instruments for measuring frequency were referenced

to the microwave transition in cesium that forms the basis for the

NIST atomic clock and contribute a negligible uncertainty. The

change in the frequency shift caused by neighboring spectral

features, a second-order effect, is negligible at the relatively low

pressures involved. The two primary components of frequency

uncertainty were combined in a root-sum-of-squares fashion

and also converted to vacuum wavelength uncertainties.

The pressure uncertainties for the high- and low-pressure

cells are estimated to be 0.38 kPa (2.9 torr) and 0.06 kPa

(0.5 torr), respectively. The uncertainty values derive from

the specified accuracy of the pressure gauge, and from the

procedures for filling and sealing the cells. After filling, the

mechanical o-ring valves of each cell were closed. In addition,

the stem of the high-pressure cell was tipped off with a torch

while the majority volume was immersed in liquid nitrogen

to condense the gas. The final pressure and uncertainty were

calculated based on the measured vapor pressure of methane

and the change in cell volume.

Uncertainties for the extrapolated values at zero pressure

were obtained with a Monte Carlo simulation using the data

from the high- and low-pressure measurements. They depend

on both the measured pressures and center frequencies (or

wavelengths) as well as their respective uncertainties. By

using the Monte Carlo technique, we were able to account for

the correlations between the different quantities. The same

technique was used to estimate the expanded (2

) uncertainty

in the pressure shift value of 1.1 MHz/kPa (0.14 MHz/torr) that

has already been mentioned.

IV. C

ONCLUSION

We have conducted a study of potential wavelength cal-

ibration references for use as moderate-accuracy transfer

standards and high-accuracy NIST internal references in the

1280–1320-nm WDM region. Among the primary considera-

tions were the availability of absorption lines, their strengths,

the ease at which they could be probed, and the material

handling issues. For our development of a NIST internal

782 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 5, MAY 2002

high-accuracy reference, we avoided a number of candidates

because they required an additional excitation step to probe

upper-level transitions. Optical, electrical, or thermal excitation

would have increased the experimental complexity and may

shift the line centers. Materials that would have been difficult

to obtain commercially or handle in the laboratory were

also avoided. Another important consideration in selecting a

material, particularly for a high-accuracy reference, was the

uncertainty to which the position of the absorption lines had

been previously measured. In the case of methane, we felt that

the strength of the spectral features and the ease of handling

justified the effort of additional measurements. Of course, the

availability of the calcium-based optical frequency standard

used in the measurements was also an important consideration.

This resource provides an in-house calibration standard at 1314

nm with an uncertainty of 13 Hz, or fractionally 6

10 .

We developed a high-accuracy wavelength reference at

1314 nm by actively stabilizing a laser to a methane line and

measuring the laser’s frequency. This methane wavelength

reference has been used in our laboratory to calibrate a wave-

length meter having an uncertainty of 1

10 . However, the

30-MHz optical width of the wavelength reference, as shown

in the mixed-down beat note of Fig. 4, can cause the last digit

of the wavelength meter readings to fluctuate. We were able

to overcome this noise, caused primarily by the dither of the

laser’s center wavelength, by averaging a sequence of readings.

A more elegant solution to this problem would be to develop

a dither-free wavelength reference using external frequency

modulation [26]. Instead of applying the dither directly to the

wavelength control of the laser, a small portion of the laser

output is passed through an external frequency modulator

driven by the dither signal. As before, the modulated light

traverses the gas cell of the reference material and produces an

error signal, which is conditioned and applied to the wavelength

control of the laser. Once stabilized, the portion of the output

that is not modulated is available for calibration purposes. We

are investigating this technique and may implement it in the

future.

EFERENCES

[1] S. L. Gilbert, W. C. Swann, and C. M. Wang, “Hydrogen cyanide

C N absorption reference for 1530–1560 nm wavelength calibra-

tion—SRM 2519,” NIST Publication 260-137, 1998.

[2] W. C. Swann and S. L. Gilbert, “Pressure-induced shift and broadening

of 1510–1540-nm acetylene wavelength calibration lines,” J. Opt. Soc.

Amer. B, vol. 17, pp. 1263–1270, July 2000.

[3] S. L. Gilbert andW. C. Swann, “Acetylene

C H absorption reference

for 1510 nm to 1540 nm wavelength calibration—SRM 2517a,” NIST

Publication 260-133, 2001.

[4] T. Dennis, W. C. Swann, and S. L. Gilbert, “Wavelength references for

1300 nm and L-band WDM,” in OSA Trends in Optics and Photonics

(TOPS) Vol. 54, Optical Fiber Communications Conference, Tech. Dig.,

Postconference ed. Washington, DC, 2001, pp. WDD83-1–WDD83-3.

[5] A. Bruner, V. Mahal, I. Kiryuschev, A. Arie, M. A. Arbore, and M. M.

Fejer, “Frequency stability at the kilohertz level of a rubidium-locked

diode laser at 192.114 THz,” Appl. Opt., vol. 37, pp. 6410–6414, Sept.

1998.

[6] H. Sasada, “Wavenumber measurements of sub-Doppler spectral lines

of Rb at 1.3

m and 1.5 m,” IEEE Photon. Technol. Lett., vol. 4, pp.

1307–1309, Nov. 1992.

[7] M. Breton, N. Cyr, P. Tremblay, M. Têtu, and R. Boucher, “Frequency

locking ofa1324 nmDFB laser to an optically pumped rubidiumvapor,”

IEEE Trans. Instrum. Meas., vol. 42, pp. 162–166, Apr. 1993.

[8] A. J. Lucero, Y. C. Chung, and R. W. Tkach, “Survey of atomic tran-

sitions for absolute frequency locking of lasers for lightwave systems,”

IEEE Photon. Technol. Lett., vol. 3, pp. 484–486, May 1991.

[9] V. Kaufman and B. Edlén, “Reference wavelengths from atomic spectra

in the range 15 Å to 25 000 Å,” J. Phys. Chem. Ref. Data, vol. 3, pp.

825–895, 1974.

[10] Y. C. Chung, “Frequency locking of a 1.3

m DFB laser using a minia-

ture argonglow lamp,” IEEE Photon. Technol. Lett., vol. 1, pp. 135–136,

June 1989.

[11] C. J. Sansonetti, “Precise measurements of hyperfine components in

the spectrum of molecular iodine,” J. Opt. Soc. Amer. B, vol. 14, pp.

1913–1920, Aug. 1997.

[12] A. Arie, M. L. Bortz, M. M. Fejer, and R. L. Byer, “Iodine spectroscopy

and absolute frequency stabilization with the second harmonic of the

1319-nm Nd :YAG laser,” Opt. Lett., vol. 18, pp. 1757–1759, Oct. 1993.

[13] L. S. Rothman, R. L. Hawkins, R. B. Wattson, and R. R. Gamache, “En-

ergy levels, intensities, and linewidths of atmospheric carbon dioxide

bands,” J. Quant. Spectrosc. Radiat. Transf., vol. 48, pp. 537–566, 1992.

[14] R. A. Toth, “Extensive measurements of H

O line frequencies and

strengths: 5750 to 7965 cm

,” Appl. Opt., vol. 33, pp. 4851–4867, July

1994.

[15] K. Chan, H. Ito, and H. Inaba, “Absorption measurement of

band of CH at 1.33 m using an InGaAsP light emitting diode,” Appl.

Opt., vol. 22, pp. 3802–3804, Dec. 1983.

[16] H. Kanamori, S. Takashima, and K. Sakurai, “Near-infrared diode

laser spectrometer with frequency calibration using internal second

harmonics,” Appl. Opt., vol. 30, pp. 3795–3798, Sept. 1991.

[17] N. Moriwaki, T. Tsuchida, Y. Takehisa, and N. Ohashi, “1.3-

m DFB

diode laser spectroscopy of

C H ,” J. Mol. Spectrosc., vol. 137, pp.

230–234, Sept. 1989.

[18] M. de Labachelerie, K. Nakagawa, Y. Awaji, and M. Ohtsu, “High-fre-

quency-stability laser at 1.5

m using Doppler-free molecular lines,”

Opt. Lett., vol. 20, pp. 572–574, Mar. 1995.

[19] E. Hirota, T. Ishiwata, K. Kawaguchi, M. Fujitake, N. Ohashi, and I.

Tanaka, “Near-infrared band of the nitrate radical NO

observed by

diode laser spectroscopy,” J. Chem. Phys., vol. 107, pp. 2829–2838,

Aug. 1997.

[20] G. Guelachvili, “Absolute wavenumber measurements of 1-0, 2-0, HF

and 2-0, H

Cl, H Cl absorption bands,” Opt. Commun., vol. 19, pp.

150–154, Oct. 1976.

[21] S. Yamaguchi and M. Suzuki, “Frequency locking of an InGaAsP semi-

conductor laser to the first overtone vibration-rotation lines of hydrogen

fluoride,” Appl. Phys. Lett., vol. 41, pp. 1034–1036, Dec. 1982.

[22] A. D. Bykov, O. V. Naumenko, M. A. Smirnov, L. N. Sinitsa, L. R.

Brown, J. Crisp, and D. Crisp, “The infrared spectrum of H

S from 1

to 5

m,” Can. J. Phys., vol. 72, pp. 989–1000, Nov.–Dec. 1994.

[23] K. Suzumura, C. Ishibashi, and H. Sasada, “Precise frequency-differ-

ence measurement between the 1.66

m transitions of methane,” Opt.

Lett., vol. 22, pp. 1356–1358, Sept. 1997.

[24] C. W. Oates, F. Bondu, R. W. Fox, and L. Hollberg, “A diode-laser op-

tical frequency standard based on laser-cooled Ca atoms: Sub-kilohertz

spectroscopy by optical shelving detection,” Eur. Phys. J. D, vol. 7, pp.

449–460, Jan. 1999.

[25] Th. Udem, S. A. Diddams, K. R. Vogel, C. W. Oates, E. A. Curtis, W. D.

Lee, W. M. Itano, R. E. Drullinger,J.C. Bergquist, and L. Hollberg,“Ab-

solute frequency measurements of the Hg

and Ca optical clock transi-

tions with a femtosecond laser,” Phys. Rev. Lett., vol. 86, pp. 4996–4999,

May 2001.

[26] G. C. Bjorklund, “Frequency-modulation spectroscopy: A new method

for measuring weak absorptions and dispersions,” Opt. Lett., vol. 5, pp.

15–17, Jan. 1980.

T. Dennis, photograph and biography not available at the time of publication.

E. A. Curtis, photograph and biography notavailable at the time of publication.

C. W. Oates, photograph and biography not available at the time of publication.

L. Hollberg (A’92), photograph and biography not available at the time of pub-

lication.

S. L. Gilbert, photograph and biography not available at the time of publication.