See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3409088

A Review of Lithium Niobate Modulators for Fiber-Optic Communications

Systems

Article in IEEE Journal of Selected Topics in Quantum Electronics · February 2000

DOI: 10.1109/2944.826874 · Source: IEEE Xplore

CITATIONS READS
1,260 5,920

11 authors, including:

Karl Kissa David Maack

Stonehill College Corning Incorporated
26 PUBLICATIONS 1,412 CITATIONS 9 PUBLICATIONS 1,277 CITATIONS

SEE PROFILE SEE PROFILE

Gregory Mcbrien
Nutmeg Pioneer LLC technical consulting
15 PUBLICATIONS 1,494 CITATIONS

SEE PROFILE

All content following this page was uploaded by David Maack on 25 December 2014.

The user has requested enhancement of the downloaded file.

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000 69

A Review of Lithium Niobate Modulators for

Fiber-Optic Communications Systems
Ed L. Wooten, Member, IEEE, Karl M. Kissa, Member, IEEE, Alfredo Yi-Yan,
Edmond J. Murphy, Senior Member, IEEE, Donald A. Lafaw, Peter F. Hallemeier, Member, IEEE,
David Maack, Member, IEEE, Daniel V. Attanasio, Daniel J. Fritz, Gregory J. McBrien, Member, IEEE,
and Donald E. Bossi, Member, IEEE

Invited Paper

Abstract—The current status of lithium-niobate external-mod- can be found in recent publications, which cover devices
ulator technology is reviewed with emphasis on design, fabrica- and system applications [1], switching technology [2], and
tion, system requirements, performance, and reliability. The tech- guided-wave devices in general [3].
nology meets the performance and reliability requirements of cur-
rent 2.5-, 10-, and 40-Gb/s digital communication systems, as well Section II of this paper describes the basic techniques used
as CATV analog systems. The current trend in device topology is to- to fabricate LiNbO3 guided-wave devices. Section III discusses
ward higher data rates and increased levels of integration. In par- various modulator device designs, structures, and functions.
ticular, multiple high-speed modulation functions, such as 10-Gb/s Section IV describes the performance of LiNbO3 devices in
return-to-zero pulse generation plus data modulation, have been digital systems. Section V covers some nondigital applications
achieved in a single device.
for LiNbO3 modulators. Section VI discusses the product
Index Terms—Electrooptical modulators, lithium niobate, op- development cycle and manufacturability of these devices.
tical modulator, waveguide devices.
Section VII reviews reliability data available from the field and
accelerated-aging tests.
I. INTRODUCTION

O VER THE past decade, as the demand for telecommu-

nications services and bandwidth has boomed, the need
for and advantages of external modulation in fiber-optic trans-
II. DEVICE FABRICATION
Lithium niobate has a very high intrinsic modulation
mission systems has been firmly established. In higher speed bandwidth, but device switching speeds are limited by a
digital communication applications, fiber dispersion has limited variety of physical constraints. Modulation is produced by a
system performance. Lithium niobate (LiNbO3) external mod- voltage-induced change in the refractive index. The achievable
ulators provide both the required bandwidth and the equally im- index change is small and, thus, either large voltages or long
portant means for minimizing the effects of dispersion. Unlike electrode lengths are needed to obtain sufficient modulation.
direct modulation of a laser diode, LiNbO3 guided-wave mod- A useful figure of merit for modulation is the product of the
ulators can be designed for zero-chirp or adjustable-chirp op- switching voltage and the electrode length. For lengths that
eration. Zero-chirp and negative-chirp modulators help to min- allow reasonable voltages, the capacitance of lumped element
imize the system degradation associated with fiber dispersion. electrodes would limit the bandwidth to less than 1 GHz. By
In analog systems, linearized external modulators can provide using travelling wave electrodes, in which the electrical signal
very low modulation distortion. propagates along the same direction as the optical wave, much
Advances in LiNbO3 modulator device technology have en- higher bandwidths can be obtained. In this case bandwidths
abled stable operation over temperature, very low bias-voltage are limited by the mismatch between the electrical and optical
drift rates, and bias-free devices. These advances in device propagation constants as well as by the electrical attenuation of
and material technology have been accompanied by significant the electrode [4]. Theoretical and experimental work has led to
investments in guided-wave device manufacturing. The net electrode structures optimized for 10-Gb/s digital modulation
result today is a strong demand for and an ample supply [5].
of high-quality LiNbO3 modulator components for use in Fig. 1 depicts a cross-sectional view of an x-cut LiNbO3
fiber-optic communication systems. In this paper, we will Mach–Zehnder interferometer (MZI) modulator, where the
provide an overview of the present state of LiNbO3 modulator critical dimensions of the modulator structure are shown.
technology. More detailed and complementary information Wave velocity is a function of the material properties and the
waveguide cross-sectional dimensions. Because the modulator
can be several centimeters long, preservation of cross-sectional
Manuscript received September 17, 1999.
The authors are with JDS Uniphase Corp., Bloomfield, CT 06002 USA. dimensions along the length of a device presents a challenge
Publisher Item Identifier S 1077-260X(00)01136-9. not only from the standpoint of fabrication but also in the
1077–260X/00$10.00 © 2000 IEEE
70 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

doping concentrations, both TE and TM modes will propagate

along the waveguides.
An alternative method for fabricating LiNbO3 waveguides
is through the annealed proton exchange (APE) process
[7]. Proton exchange [8] is a low-temperature process
(∼120 C–250 C) whereby Li ions from the LiNbO3 wafer are
exchanged with protons from an acid bath. Proton exchange
can only be applied to x- and z-cut wafers; the acid chemically
etches y-cut wafers. Exchanged layers exhibit an increase in the
Fig. 1. Cross-sectional view of an x-cut modulator with coplanar-waveguide
(CPW) radio-frequency electrodes. extraordinary index; the ordinary index is left virtually undis-
turbed or even reduced. APE waveguides are, consequently,
polarizing waveguides. The proton exchange is followed by a
choice and compatibility of materials used. Required materials high-temperature anneal. The annealing step in the APE process
include the electrooptic substrate, electrode metal, electrode is crucial in obtaining high-quality electrooptic waveguides.
adhesion layer, buffer layer, and the dopant used for fabrication The annealing process successfully addresses the problems of
of the optical waveguides. index instability [9], [10] and deterioration of the electrooptic
effect [11] exhibited by proton-exchanged waveguides without
A. LiNbO3 Wafers anneal. Accelerated aging studies carried out on x-cut APE
directional couplers and Mach–Zehnder modulators have
LiNbO3 has been the material of choice for the fabrication
shown no degradation of device performance over a 13-year
of electrooptic modulators due to its combination of high elec-
lifetime at 125 C and a 25-year lifetime at 95 C [12], [13].
trooptic coefficients and high optical transparency in the near in-
Preservation of waveguide cross-sectional properties across
frared wavelengths used for telecommunications. Its high Curie
the wafer requires good uniformity in Ti film deposition as well
temperature (1100 C–1180 C) makes it practical for fabri-
as uniformly flat thermal zones in furnaces used for the diffusion
cation of low-loss optical waveguides through indiffusion of
or exchange/anneal processes. Careful screening of wafers for
metals [6]. LiNbO3 is also thermally, chemically, and mechani-
surface flatness is key to minimizing waveguide-width variation
cally stable and is compatible with conventional integrated-cir-
across the wafer during mask definition by photolithography.
cuit processing technology.
Last, process parameters such as Ti film thickness, diffusion
Wafers with high optical homogeneity and surface flatness
temperature, and time can be tailored to yield high-quality stripe
are needed in the fabrication of integrated optical devices.
waveguides, which exhibit low propagation loss, optimum field
LiNbO3 wafers are obtained from boules grown using the
confinement, and good waveguide-to-fiber coupling.
Czochralski technique. During the growth process, care is taken
Almost all commercially available 2.5- and 10-Gb/s LiNbO3
to ensure that Fe ion contaminants are eliminated, or kept to a
modulators use stripe optical waveguides. As modulation speeds
minimum, in order to reduce the susceptibility of the crystal to
increase to greater than 20 Gb/s, ridge waveguides on z-cut sub-
photorefractive optical damage. LiNbO3 is particularly suscep-
strates will be required to achieve lower drive voltages and re-
tible to optical damage at wavelengths below 1 m. However,
duce device lengths [14]. A ridge waveguide under the active
for wavelengths greater than 1 m, photorefractive damage is
(or “hot”) electrode concentrates the flux of the applied electric
generally negligible for optical powers less than 100 mW.
field within the waveguide and provides tighter confinement of
LiNbO3 wafers with diameters up to 100 mm are commer-
the optical mode. Both effects increase the overlap between the
cially available in different crystal cuts (x-, y-, and z-cut) with
applied electric field and the optical mode in a z-cut arrange-
the choice of cut depending upon the application. The handling
ment, thus providing a lower switching voltage. LiNbO3 ridges
of x- and y-cut wafers during the fabrication process is fairly
a few micrometers deep can be obtained by wet etching [15],
straightforward, whereas z-cut wafers require special handling
[16] or ion-milling [17] techniques. Smooth ridge sidewalls as
procedures because the crystal’s piezoelectric and pyroelectric
well as good control of etch depth and sidewall direction are key
properties produce electrical charge accumulation on the z-faces
in achieving low-loss waveguides with good mode confinement.
during processing. Modulators made on z-cut wafers also re-
quire special design and packaging to minimize bias drift due to
charge migration and the buildup of pyroelectric charges. C. Electrode Fabrication
RF electrodes are fabricated either directly on the surface of
B. Waveguide Fabrication
the LiNbO3 wafer or on an optically transparent buffer layer to
LiNbO3 waveguides have been traditionally fabricated by in- reduce optical loss due to metal loading and provide a means for
diffusion of Ti at temperatures near 1000 C. The process is optical/RF velocity matching. In general an adhesion layer, such
straightforward, but precautions must be taken to suppress the as Ti, is first vacuum deposited on the wafer, followed by the
out-diffusion of Li from the surface of the substrate. A Li-defi- deposition of a base layer of the metal in which the electrodes
cient surface results in an unwanted planar waveguide for z-po- are to be made. The electrode pattern is then photolithographi-
larized light and can seriously affect modulator performance. cally defined. High-speed modulators incorporate RF electrodes
The presence of Ti in the crystal increases both the ordinary whose thickness can range from a few micrometers to greater
and the extraordinary indexes of refraction. Hence, for suitable than 15 m. Special lithographic processes are needed to define
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 71

tively directed into and away from the device. Three principle
subassemblies or subcomponents are used in the manufacture
of packaged LiNbO3 modulators. The three subassemblies are,
namely, the integrated-optic chip, optical-fiber assemblies, and
electrical or RF interconnects and housing. The fabrication of
the integrated-optic chip was described in detail in the preceding
paragraphs of this article.
LiNbO3 modulators can be packaged in either hermetic or
nonhermetic housings, depending upon the application, oper-
ating environment, and methods and materials utilized in the
modulator manufacturing process. Reference [18] provides a
comprehensive description of the packaging techniques used
to produce hermetically packaged LiNbO3 modulators. For de-
vices designed to operate within telecommunication central-of-
fice environments, nonhermetic packaging has proven both suf-
ficient and cost effective for meeting the necessary reliability
and qualification requirements [19].
Due to the polarization dependence of the electrooptic effect,
Fig. 2. SEM picture of 18-m-thick gold-plated CPW electrodes at a region the polarization state of the input light supplied to the modu-
where the electrodes are making a bend.
lator must be carefully controlled and maintained to achieve
optimum performance. Hence, most LiNbO3 modulators uti-
the thick mask and to ensure that the electroplating process pre- lize polarization-maintaining fiber for the input pigtail, while
serves the cross-section of the RF electrode along the length of the output fiber is typically standard single-mode fiber. During
the device. the preparation of the input and output optical fiber subassem-
Gold is generally used as the electrode metal. Plating pro- blies, a small tube or block is attached to the end of each fiber.
cesses yielding high-purity metal, small grain size, minimum This tube is angle cut and polished to minimize back-reflections
feature distortion, and reasonable plating rate are key to from the fiber-to-LiNbO3 interface. The tube also increases the
obtaining good RF performance. Various plating chemistries, surface area and bond strength of the pigtail joint, thereby en-
dry or in solution, are commercially available. Each chemistry suring long-term stability and reliability. Ultraviolet-curing or
works best in conjunction with a particular continuous and/or thermal adhesives are used to attach the fiber subassemblies to
pulsed current supply as well as a particular mask material. the LiNbO3 chip. The fiber subassemblies are typically designed
After plating, the mask is removed and the metal in the gaps with a strain-absorbing bow to accommodate the differential
is etched away. Fig. 2 shows a scanning electron microscope thermal expansion among the LiNbO3 chip, the fiber, and the
(SEM) picture of a typical gold-plated coplanar-waveguide metallic housing [18].
electrode fabricated for a 10-Gb/s digital modulator. Electrical interconnects are attached and soldered to the mod-
ulator housing, thereby creating the third subassembly. The pig-
D. Dicing and Polishing tailed LiNbO3 chip is attached to the package using a compliant
adhesive that mechanically decouples the optical assembly from
In contrast to semiconductor materials such as GaAs or InP, the package and absorbs thermally induced strains [20]. Last,
LiNbO3 substrates do not cleave readily. Substrates containing electrical interconnection between the package and the LiNbO3
an array of finished modulators are cut from the LiNbO3 wafer chip is accomplished using either wire or ribbon bonds. For
using conventional water-cooled diamond saws. The substrate high-frequency operation, the dimensions of the bond must be
end faces are cut at an angle to the waveguides in order to elimi- carefully controlled to achieve optimal performance.
nate reflections and are then polished to an optical finish. A good Once all three subassemblies are copackaged, the modulator
optical finish and a sharp edge are required at both the input and must be sealed and tested as a completed, functional unit. Key
output optical facets of the device to ensure good fiber-to-wave- parameters that are typically measured during final test include:
guide coupling. For ease of handling, the modulators on a sub- optical loss, switching voltage, optical on/off extinction,
strate are individually tested before the substrate is diced into bias stability, and microwave/RF properties such as and
individual components. . Oftentimes, some of these performance parameters are
Cleanliness must be maintained throughout the modulator measured over the operating temperature range of the device.
fabrication process. Debris from dicing and particulates from Bit-error-rate and eye-diagram measurements are not typically
polishing compounds are contaminants that can negatively im- required on each production device. These measurements are
pact the performance and long-term reliability of the modulators, often performed during the design verification and validation
and must therefore be removed during chip cleaning operations. stages of product development, and only the aforementioned
parameters need to be measured in production.
E. Pigtailing, Packaging, and Test
III. DEVICE DESIGN
To utilize LiNbO3 modulators in real-world applications, the
integrated-optic chip must be pigtailed and packaged in order The basic building blocks for lithium niobate modulators and
that optical and electrical signals can be efficiently and effec- switches generally fall into one of two categories: MZI type or
72 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

directional coupler type [21]. In the MZI, light is split into two
optical paths that are isolated from one another. The applied
electric field from an electrode modifies the relative velocities
of the two beams via the electrooptic effect, resulting in variable
interference when the two paths are recombined at the output.
In the directional coupler approach, light is injected into two
modes of a waveguide structure. The applied field modifies the
relative velocities of the two modes as well as the coupling be-
tween the modes. The two most common examples of the direc-
tional coupler type are the reversed delta- coupler [22] and the
digital optical switch [23]–[27]. The reversed delta- coupler
is compact and can be tuned with modest voltages (10–20 V).
The digital optical switch requires higher drive voltages (40–50
V) but can be constructed as a polarization-independent switch.
The MZI works well with high-bandwidth electrode structures
requiring tens of micrometers of spacing between waveguides,
and long electrodes are needed to reduce drive voltages (
V). The directional-coupler-type switches are typically used for
lower speed switching applications where small size and polar-
ization diversity may be required, and tight electrode gaps (<10
m) are more easily accommodated.
The first choice encountered in designing a LiNbO3 modu-
lator is the orientation of the crystal axes to the waveguides and
electrodes. The crystal cut affects both modulator efficiency, as
denoted by half-wave voltage , and modulator chirp, which is
described by the chirp parameter . The difference is apparent in
Fig. 3, which shows the four most common electrode structures
used in MZI-type switches. The strongest component of the ap-
plied electric field must be aligned with the -axis of the crystal,
which has the highest electrooptic coefficient. This requires that
the waveguide be placed between the electrodes for an x-cut
configuration and beneath the electrodes for z-cut. Because the
electrodes are placed on top of the waveguides, z-cut devices
always require a buffer layer to minimize attenuation of the op-
tical mode due to metal absorption. Z-cut devices also typically
employ conductive buffer layers and charge bleed layers to mit-
igate dc drift and pyroelectric charge buildup, respectively [18].
X-cut devices do not inherently need a buffer layer because the
electrodes are not placed directly above the waveguides; how-
ever, to achieve multigigahertz operation, broad-band x-cut de-
vices do use a buffer layer for velocity matching of the RF and
optical waves.
The applied electric field and electrooptic efficiency of
various electrode topologies can be modeled using quasi-static Fig. 3. Most common electrode configurations for (a) nonbuffered x-cut, (b)
techniques such as finite-element or finite-difference methods buffered x-cut, (c) buffered single-drive z-cut, and (d) buffered dual drive z-cut.
[28]. These techniques also provide the microwave properties
of the electrode (velocity, impedance, and loss). RF loss applied fields in the electrode gaps. In z-cut devices, the
can also be adequately determined at high frequencies (>2 waveguide positioned underneath the hot electrode experiences
GHz) with Wheeler’s inductance rule [29]. In general, thick an RF field flux that is more concentrated, resulting in a factor
electro-plated electrodes ( m) have low RF loss ( of two improvement in overlap between RF and optical field,
dB cm−1 GHz−0.5) and enhanced velocity matching due to the relative to x-cut. However, the overlap under the z-cut ground
presence of electric flux in the air gap between electrodes. electrode is reduced by a factor of three, relative to x-cut;
Buffer layers are required for broad-band velocity matching therefore, the overall improvement in z-cut (relative to the
on both x- and z-cut devices due to the high RF dielectric x-cut ) is only about 20% for single-drive modulators. The
constants of lithium niobate ( ) relative to the difference in overlap between the two z-cut waveguides results
optical dielectric constants ( ). in a chirp parameter of approximately −0.7.
X-cut electrode topologies [Fig. 3(a) and (b)] result in By employing a dual-drive topology, in which the push–pull
chirp-free modulation due to the push–pull symmetry of the effect is produced by the driver circuit, the factor of two im-
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 73

Fig. 4. Calculated electrooptic response for several devices designed Fig. 5. Calculated electrooptic response for the case where all modulators have
for broad-band digital or RZ pulsing applications at 2.5 or 10 Gb/s a 5-cm electrode length (conventional electrode and waveguide structures, no
using conventional electrode and waveguide structures (no ridges or n 1 1
ridges or n enhancement).
enhancement). Electrode lengths were chosen to be representative of devices
used in actual applications and are shown in Table I.
TABLE I
NORMALIZED ELECTRODE LENGTHS USED
provement in overlap under the hot electrode can be utilized in IN CALCULATION SHOWN IN Fig. 4

both arms of the MZI at the expense of increased RF circuit

complexity. Balancing the two drive levels results in zero-chirp
operation. Imbalancing the two drive levels provides electroni-
cally programmable chirp.
Nonbuffered x-cut is the overall winner in electrooptic effi-
ciency per unit length, having about a 30% lower than the
z-cut waveguide beneath the hot electrode. Unfortunately, ve-
locity mismatch of nonbuffered x-cut limits bandwidth, and the
25- electrode impedance introduces additional penalties due Narrow-band devices are well suited for return-to-zero (RZ)
to impedance mismatch and RF loss. RF electrodes on buffered pulsing applications because the pulse generator is typically
x- and z-cut substrates can be designed for velocity matching driven with a sinusoidal signal at the clock frequency [34].
and impedances near 40 . The electrooptic efficiency versus modulation frequency is
A number of methods have been employed to enhance the compared in Figs. 4 and 5 for the various electrode geometries
electrooptic efficiency. Ridge waveguides further enhance the and crystal cuts of Fig. 3. A narrow-band x-cut modulator for
focusing of electric flux under the hot electrode and have been RZ pulsing is included as well [33]. The response curves rep-
used to demonstrate of 3.5 and 5.1 V for z-cut modulators resent electrical power at the receiver’s photodetector per unit
with 30- and 70-GHz bandwidths (3 dB electrical), respectively of total electrical power into the modulator (add the power into
[28]. A 15% reduction of has also been demonstrated by both ports for dual drive). Conventional electrode and wave-
using ridge structures in an x-cut device without a buffer layer guide structures are assumed [28], [35], that is, CPW electrodes
and by using lumped-element (low-frequency) electrodes [30]. are assumed and no ridges or enhancement are utilized. RF
The overlap between the applied electric field and the optical properties and RF-optical overlap have been calculated using fi-
fields can also be improved in both x- and z-cut by increasing the nite difference quasi-static field analysis. Optical mode profiles
waveguide refractive index and narrowing the ground electrode typical of conventional single-mode stripe wave-guides are used
width [31]. in the RF-optical field overlap calculation. The electrode prop-
The electrooptic efficiency at modulation frequency can erties and dc electrooptic efficiency are inserted into an elec-
be improved at the expense of bandwidth by using RF trooptic response model that accounts for reflections from the
phase-matching techniques such as phase reversal or intermit- RF termination [35].
tent-interaction electrodes [32]. These narrow-band techniques Fig. 4 shows the calculated electrooptic response for several
allow the buffer/electrode structure to be optimized for the devices designed for broad-band digital or RZ pulsing transmis-
best field overlap, while RF phase-matching is achieved by sion applications at 2.5 or 10 Gb/s using conventional electrode
flipping the polarity of the electrooptic interaction periodically, and waveguide structures (no ridges or enhancement). The
or by moving the electrode away from the waveguide when electrode lengths were chosen to be representative of devices
the walk-off between RF and optical fields exceeds 180 . used in actual applications and are shown in Table I, normal-
More complex electrode structures allow further improvement ized to the shortest electrode (nonbuffered x-cut). The applica-
in phase-matching while still maintaining the high level of tion for which each device would be best suited is also listed
field overlap, resulting in approximately 40% improvement in Table I. Fig. 5 shows the electrooptic response for the case
in modulation efficiency compared to phase reversal [33]. where all modulators have a 5-cm electrode length. Note that
74 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

the plots in Figs. 4 and 5 are of the same format as an (elec- electrode structures, to form a chirped RZ pulse generator [33]
trooptic electrical dB) network analyzer measurement. RF loss needed for long-distance transmission [34], [37]. Integration of
from packaging is neglected in the plots of Figs. 4 and 5, in order MZI’s in series or parallel has been proposed and demonstrated
that raw electrooptic performance can be compared. for linearization of the transfer function for cable TV and
The comparison of electrooptic efficiency in Fig. 4 reveals other analog transmission applications [38]–[40]. Last, hybrid
some interesting results. As expected, z-cut dual drive is the integration of photodetectors with the lithium niobate chip is
most efficient, given its long length, velocity matching, and readily accomplished for power monitoring functions.
high overlap efficiency. The 4-dB advantage over buffered x-cut
is mainly due to the factor-of-two improvement in RF-optical
IV. SYSTEMS REQUIREMENTS AND DIGITAL PERFORMANCE
overlap minus the 3-dB penalty for two inputs. Single-drive
z-cut only boasts a 2-dB advantage over buffered x-cut, and In recent years, LiNbO3 devices have successfully addressed
is surpassed slightly at 10 GHz by narrow-band x-cut, which the modulation requirements in digital fiber-optic time-do-
is 30% shorter. Nonbuffered x-cut begins near the efficiency main-multiplexed (TDM) and wavelength-division-multiplexed
of z-cut single drive, but rolls off faster due to RF-optical ve- (WDM) systems. Modulators have become a critical com-
locity mismatch. Note that the improved performance of z-cut ponent in the majority of today’s long-haul-terrestrial and
dual-drive device comes at the expense of greater sophistication submarine optical networks. The appropriate modulator archi-
in delivery of RF drive signals to the modulator, a complication tecture is chosen for a given system implementation based on
that is an issue at 10 Gb/s and becomes problematic at higher performance criteria and transmission format.
bit rates such as 40 Gb/s. When designing and characterizing a fiber-optic transmis-
The electrooptic efficiency for the case where all electrodes sion system, a baseline measurement of the transmitted signal
are 5 cm is shown in Fig. 5. The narrow-band x- cut device is the is made at the modulator output through analysis of the signal’s
overall winner at 10 GHz, even surpassing the z-cut dual-drive time-domain properties. This modulated optical signal is then
structure by about 1 dB. The low RF electrode loss and high launched into the fiber-optic link. Signal degradation, resulting
overlap efficiency of the narrow-band electrodes on nonbuffered from numerous fiber processes, amplifier noise, and other link
x-cut substrates account for the improved performance. In the components, is quantified by measuring the detected signal. A
narrow-band design, the RF loss can be minimized with little measurement of the bit error rate (BER) determines the accu-
constraint from velocity or impedance-matching concerns, a de- racy with which the optical receiver is able to differentiate be-
gree of freedom that becomes significant for longer electrode tween digital “one” and “zero” bits. This fundamental challenge
lengths. of maintaining signal quality through the optical fiber link de-
One other factor that affects modulator design is bias voltage termines the performance requirements of the modulator.
drift. For MZI type modulators, the optical output power versus The predominant technique used for telecommunications
drive voltage transfer function is sinusoidal, with the ideal bias transmission in optical fiber has been on–off keying (OOK), or
point near the half power point (quadrature). Quality of trans- intensity modulation with direct detection, where the optical
mitted digital data can suffer if the bias point shifts too much intensity is gated in a shuttering fashion. Data encoding has
over time, generally requiring an active feedback circuit to mon- predominantly utilized a nonreturn-to-zero (NRZ) format,
itor and adjust the bias voltage. Alternatively, the bias point where an arbitrary data stream of “one” and “zero” symbols is
can be preset in manufacturing, eliminating the need for a feed- directly encoded as high and low optical powers, respectively.
back circuit, or at the very least reducing the level of applied dc Advantages of OOK include minimizing the fiber link degrada-
voltage. The long-term dc drift behavior is best described by an tion due to group velocity dispersion, because NRZ data coding
RC ladder model [36]. The initial applied field is determined by provides a narrow-spectrum signal, and ease of data recovery at
capacitive voltage division through the ladder. The long-term the receiver. OOK is adequate for links operating at data rates
field is determined by resistive voltage division. The resistors of 2.5 Gb/s for spans up to 1000 km and 10 Gb/s for spans up
and capacitors in the model are set by process parameters and to 500 km, utilizing traditional and nonzero-dispersion-shifted
conditions used to fabricate the waveguides and electrodes. Ac- single-mode fibers.
celerated aging tests reveal that bias voltage for a device using A typical NRZ transmitter incorporating a LiNbO3 modulator
Ti indiffused waveguides will not change by more than a factor is shown in Fig. 6. Digital OOK waveforms are typically pre-
of two from the initial bias voltage, over a 20-year lifetime at sented in the form of an “eye” diagram, which is the superpo-
typical operating temperatures. sition of all “one” and “zero” states of a pseudorandom bit se-
Lithium niobate modulator technology easily accommodates quences (PRBS) within one bit window. Fig. 7(a) and (b) shows
integration of multiple functions. The level of integration is LiNbO3 externally modulated eyes at 2.5 and 10 Gb/s, respec-
mainly limited by the size of the wafer and the length of the tively.
different building blocks. A simple example of integration is The center of the eye is required to receive no light intensity,
a series cascade of a low-speed MZI and a high-speed MZI or “hits,” since the receiver decision circuitry operates in this
to form a variable optical attenuator plus data modulator. The region using a preset intensity threshold. A larger “no-hit” zone
efficiency per unit length of the narrow-band x-cut electrode improves the accuracy of the receiver and lowers the BER. Since
permits an RZ pulse generator to be integrated in series with the processes that degrade the eye are typically stochastic in na-
a broad-band data modulator [33]. Alternatively, an MZI and ture, long sampling times of the transmitted signals are required
phase modulator can be integrated, both having narrow-band to accurately quantify the system’s BER. Transmitter metrics,
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 75

of the modulator that impact these metrics, and therefore de-

termine system performance, are switching voltage, elec-
trical-to-optical response, electrical return loss, optical ex-
tinction ratio, excess optical loss, and optical chirp of the mod-
ulated signal.
The parameter relates the modulator’s high-frequency
switching voltage to the electrooptic efficiency of the device at
low frequency, usually denoted as at dc. Practical implemen-
Fig. 6. Data modulator topology for NRZ transmission. tations of digital systems can tolerate up to 3-dB rolloff in effi-
ciency at the highest frequency in the modulated signal’s spec-
trum. In digital systems, the requirement for wide-band opera-
tion is driven by the nature of the data signals, which are, by def-
inition, random with continually varying frequency content. As
an example, for 2.5-Gb/s operation, the 3-dB bandwidth of the
response is normally specified to be greater than 2.2 GHz.
The phase linearity of the electrooptic response, describing
the linear phase offset error during the electrical-to-optical con-
version at a particular frequency, must be less than ± 5 over
the operating frequency band. The parameter, or the elec-
trical return loss at the modulator drive port, is also of great im-
portance. The value must typically be less than −10 dB, in
order that the reflected power returning to the high-speed drive
electronics is low and that there is <5% reduction in electrooptic
conversion efficiency.
The optical extinction ratio describes the ability of the device
to extinguish the light when the modulator is driven to a “zero”
state. The extinction ratio is typically required to be greater than
20 dB at dc for digital applications. Variation in the modulator’s
electrooptic response with frequency and some small amount of
reflected RF power between the modulator and driver reduce
the observed extinction ratio for a PRBS signal. The optical
“eye” extinction is one of the most important figures of merit
(a) for the signal, especially when cascading a number of link sec-
tions and optical amplifiers. From [41], the expression for link
power penalty as a function of extinction ratio is

where is the linear ratio of ON power to OFF power. An in-

crease in extinction ratio from 10 dB ( ) to an extinc-
tion ratio of 20 dB ( ) results in ∼0.8-dB improve-
ment in receiver sensitivity. For appropriately designed trans-
mitters utilizing LiNbO3 modulators, 20 dB extinction at fre-
quency and >90% eye opening has been demonstrated at 2.5
Gb/s. These transmitters readily maintain >15 dB extinction
under all operating conditions. At 10 Gb/s, >15 dB extinction
has been achieved at frequency and >13 dB maintained over all
conditions.
Modulator chirp must be taken into consideration in the link
design, due to its interaction with dispersive effects. One of the
(b)
major strengths of LiNbO3 modulators is that they can be de-
Fig. 7. LiNbO3 externally modulated eyes at (a) 2.5 and (b) 10 Gb/s. signed to have zero chirp and suffer no link length penalty due to
dispersive interaction. However, to solve some of today’s higher
including average launch power, optical extinction ratio, tran- bit-rate problems, the ability to design a nonzero-chirp LiNbO3
sition rise and fall times, timing jitter, transition crossing per- modulator has also proved important. The design of the mod-
centage, and parameter, have been developed to quantify the ulator’s chirp value is used as a degree of freedom, which can
quality of the digital signal. The key performance parameters be leveraged to extend link distances. Chirp has been shown to
76 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

TABLE II
MODULATION FORMATS

Fig. 8. Pulse-data modulator topology for RZ transmission.

enhance the transmission distances at 10 Gb/s by 25% (from 80 Hence, modulator suppliers have designed multistage LiNbO3
to 100 km) [42]. devices, which ease these requirements. Specifically, the
The very high DWDM channel loading (100 channels or bandwidth requirement for the higher spectral-width RZ format
more), the increased bit-rate requirements of next-generation is divided into two optical stages, each with lower frequency
systems, and the desire to build wavelength-intelligent net- drives. Fig. 8 shows a RZ transmitter topology implemented
works has pushed the capabilities of the NRZ transmission to using a two-stage LiNbO3 device. The output of such a RZ
its limits. Hence, other data-encoding formats and modulation transmitter is shown in Fig. 9. NRZ input signal and RZ
techniques, as described in Table II, are being considered to transmitted data at 10 Gb/s are shown in Fig. 9. These RZ
improve system performance. Lithium niobate affords the transmitters have also been extended to include a third stage,
advantages of a customizable modulator device technology which provides phase modulation, used to synchronously chirp
that can be used to supply enabling building blocks for these the pulse train and extend even further the transmission link
alternative modulation formats and techniques. span.
The return-to-zero modulation format has been employed The ability to integrate multiple functions into a single com-
in recent high-bandwidth terrestrial and submarine systems, ponent has been a powerful attribute of LiNbO3 modulators,
especially those requiring long transmission distances. Dis- driven by the desire to solve system-related challenges using the
persion managed soliton and other narrow-pulse transmission capabilities of the LiNbO3 chip. Multifunction LiNbO3 compo-
techniques can be considered specialized versions of RZ trans- nents can be realized with a variety of integrated-optical struc-
mission. In conventional NRZ pulse format, the interaction tures connected in series, parallel, or a combination of both.
between self-phase modulation (SPM) and group velocity Parallel-integration approaches are attractive in areas of
dispersion (GVD) causes the transfer of energy from the center the network where signals pass through multiplexers and
of the pulse toward the pulse edges [43], [44], a situation that demultiplexers. In these cases, many wavelengths, propagating
is enhanced for a stream of “ones.” Cross-phase modulation in separate fibers, are present on a single card and arrayed
(XPM) worsens for a stream of “ones,” as well, due to the devices (the integration of multiple-parallel channel paths on
increased cross-wavelength interaction time. Use of the RZ a single chip) are particularly useful. Applications such as
format in a dispersion-managed system greatly reduces SPM photonic switching [45], channel-to-channel attenuation, and
and GVD pulse deformation, resulting in greater pulse-to-pulse 3-R regeneration within line equipment are some of the drivers
consistency. The improved performance permits lower optical for parallel-integration technology.
power levels, due to the reduced duty cycle, which in itself In other applications, multiple component functions are inte-
provides additional relief from the nonlinear effects. grated in series. For example, an integrated phase modulator and
RZ modulation increases the demands on bandwidth and amplitude modulator can be used to provide adjustable chirp to
phase linearity for both the modulator and the drive electronics. the modulated signal launched into a network, allowing for ex-
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 77

Fig. 9. Acquired waveforms of 10-Gb/s NRZ and RZ data.

tended reach in networks using dispersive fiber. Another prac- nondigital communications applications, such as CATV signal
tical example of serially combining functions is the integra- distribution and wide-band microwave signal distribution and
tion of the optical data modulator with an on-chip variable at- antenna remoting.
tenuator. Today’s DWDM systems require channel-to-channel It is generally accepted that high-quality, “head-end” CATV
variable attenuation for dynamic power equalization [41] and signal distribution (typically, 40–860 MHz) was the first truly
full channel blocking during live-card installations (hot swaps). successful commercial application of LiNbO3 external modu-
Today, several LiNbO3 component suppliers are integrating the lator technology. State-of-the-art CATV modulators are special-
attenuation and modulation functions into a single package. ized, high-performance devices. They typically include an am-
Future-generation wavelength-agile networks will require plitude modulator, a phase modulator for suppression of stimu-
modulation components to be wavelength independent and to lated Brillouin scattering, a separate bias port, and two comple-
operate in wavelength bands outside the traditional C-band of mentary optical outputs. The amplitude modulator must have a
1530–1565 nm. New systems are encompassing the L-band flat electrooptic ( ) response (±0.5 dB) and low electrical re-
of 1570–1610 nm. Lithium niobate modulators have been turn loss ( < −18 dB). An extensive review of external mod-
designed to maintain low loss and constant switching voltage ulation for CATV signal distribution can be found in [48].
over the 1530–1610-nm wavelength range. As optical-amplifier Traditionally, electronic predistortion of the CATV electrical
technology matures, new markets will develop for modulators signal has been required in order to limit the inherent nonlinear,
and other optical components at many wavelengths. Lithium odd-order distortion products arising from the optical modu-
niobate’s broad wavelength flexibility will be invaluable in con- lator [49]. The development and successful commercialization
struction of these devices. As an example, future networks may of optically linearized modulators has virtually eliminated the
use tunable lasers to provide wavelength-selectable transmis- need for such electronic predistortion and its associated finely
sion sources to add functionality, such as wavelength routing, tuned predistortion circuitry [50]. Either method can be used
and to help solve sparing and inventory issues [47]. Tunable to transmit signals with low distortion and high carrier-to-noise
laser transmitters require the use of an external modulator ratio.
in order to provide a high-quality transmission signal, while One example of an optically linearized modulator uses two
maintaining wavelength-stabile operation. The modulator must independent amplitude modulators connected in series via a di-
have a broad-band response to wavelength in order to provide rectional coupler. A second directional coupler is required at the
consistent performance over the entire tuning range of the laser. output of the second amplitude modulator. The RF input (am-
Laboratory demonstrations of tunable lasers, modulated with plitude and phase) to each of the two series modulators is con-
LiNbO3 modulators, have been successfully performed. trolled separately. In order to achieve adequate reduction of the
distortion product, the linearized modulator requires that the two
V. NONDIGITAL APPLICATIONS directional couplers be fabricated with coupling angles within
In the environment of the late 1990’s, where WDM-based 2% (i.e., 4 ) of one another. This represents a fairly tight, yet
digital systems are being deployed in high volume, it is easy achievable, manufacturing tolerance.
to overlook the equally important role of LiNbO3 modulators in The flat attenuation-versus-frequency characteristic of
nondigital applications. Commercially available LiNbO3 mod- optical fiber makes it an ideal medium for microwave and
ulators are today being deployed in a variety of lower volume millimeter-wave transmission, as opposed to lossy coaxial
78 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

cable or heavy metallic waveguides. For this reason, high-speed

external modulators have been the object of intense research
and development since the inception of LiNbO3 guided-wave
devices. In the last few years, packaged LiNbO3 devices with
bandwidths in excess of 40 GHz [51] have been demonstrated.
Responses up to 100 GHz have been measured in a probed
device [52].
In spite of these recent advances, the deployment of true mi-
crowave frequency links has not been widespread. They have
typically been limited to military and aerospace applications
where multioctave bandwidths and long-distance transmission
for microwave frequencies are required [53]. These applications
include radar delay lines, towed aircraft decoys and wide-band
antenna remoting for distances greater than 100 m.
The widespread acceptance of multigigahertz links has been
hampered by the relatively high drive voltages required by com-
mercially available modulators, typically in the 6–10 V range at
10 GHz, and even larger at higher frequencies. Modulator drive
voltages of this magnitude require signal amplification in the
35–45 dB range, assuming about 2–3 mA of detector current.
Unlike digital systems whose wide-bandwidth signal is carried
directly as generated, RF distribution systems have the option of
down-conversion to lower frequencies, which can be transported Fig. 10. The generic sequence of steps required to produce LiNbO3
with much lower losses on standard coaxial cables. For LiNbO3 electrooptic modulators.
to achieve widespread deployment in wide-band, multigigahertz
analog systems, a truly transparent link (0 dB electrical loss) is be categorized into three general groups. These are integrated
needed. This requires an increase in modulation efficiency by optical circuit fabrication, device packaging, and testing. Each
one or two orders of magnitude. A detailed treatment of inten- of these can be further categorized. For example, integrated
sity modulation in analog fiber-optic links may be found in [54]. optical circuit fabrication consists of waveguide processes,
buffer layer processes, electrode processes, etc. Device pack-
VI. PRODUCT DEVELOPMENT AND MANUFACTURABILITY aging consists of RF interconnection, optical interconnection,
soldering and sealing processes, etc. Testing consists of
Like all manufactured products, the manufacturability of in-process testing, wafer-level testing, final device testing, etc.
LiNbO3 electrooptic modulators is established in the product The generic sequence of steps required to produce LiNbO3
design and development cycle. The job of the product devel- electrooptic modulators is shown in Fig. 10.
opment team is to anticipate and understand the manufacturing Standardization within each process across different prod-
issues in order to generate product designs that meet high ucts, reduces performance variability, simplifies material
performance requirements and that can be built in volume requirements, reduces development time, and ultimately yields
on the production floor. The primary challenge in realizing a more manufacturable products in production. Through solid
manufacturable design during the product-development process design rules and standardization of processes, LiNbO3 mod-
stems from the tradeoffs among performance, “standard” man- ulator technology has rapidly transitioned from a laboratory
ufacturing process capability, and development time. research topic to high-volume manufacturing. Over the past five
The importance of “time to market” in the photonic telecom- years, more than 100 000 LiNbO3 modulator devices have been
munications industry cannot be underestimated. Because the deployed in commercial fiber-optic communication systems.
industry is moving at an extremely rapid pace, product-devel-
opment cycle time is as critical as performance. This is espe-
VII. RELIABILITY
cially true for photonic components, like LiNbO3 electrooptic
modulators. System engineers need components that meet per- An important device performance parameter for LiNbO3
formance specifications, but they also need components that modulators is field reliability. Adequate field data are be-
are available in the time frame of their system design window. coming available to confirm and corroborate qualification and
In addition, the device design must be able to rapidly transi- long-term reliability predictions. The field results demonstrate
tion to high-volume production to meet the demands of today’s that stability; performance and reliability have reached levels
fast-paced communications marketplace. only speculated of a few years ago. Table III shows actual
Modular designs that use standard design elements and field experience for two of the more common types of LiNbO3
standard manufacturing processes enable the reduction of modulators.
product-development cycle times, while simultaneously en- The calculated failure rates in Table III assume that field fail-
suring manufacturability. Standard manufacturing processes ures are random and are best described by an exponential dis-
used in the production of LiNbO3 electrooptic modulators can tribution. The confidence intervals are calculated using a chi-
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 79

TABLE III
FIELD FAILURE RATES FOR LiNbO3 MODULATORS

(a)

(a)
(b)
Fig. 12. Phase drift for 2.5 Gb/s bias free, x-cut, APE modulators at (a) 85 and
(b) 100 C [19].

(b)
Fig. 11. Bias voltage drift for 2.5 Gb/s, x-cut, Ti modulators at (a) 85 and (b)
100 C [19].

squared estimator. FIT stands for failures in time per 10 device

hours, i.e., parts per billion. As can be seen from the table, the
field failure rates for the listed devices are extremely low.
One of the largest concerns during the development of
LiNbO3 technology was the issue of bias drift. Along with the Fig. 13. Bias voltage drift at various temperatures, 10 Gb/s, z-cut, Ti
availability of a significant amount of field data, the results modulators, from Minford [55].
from long-term accelerated aging studies are also becoming
available. The results strongly suggest that bias drift is not
a failure mechanism in properly designed and manufactured
devices. Figs. 11–14 show the results of bias drift studies on a
number of device types from different manufacturers. Clearly,
all of these curves show excellent behavior.
Fig. 11(a) and (b) shows the drift of x-cut 2.5-Gb/s tita-
nium-diffused waveguide devices at 85 and 100 C, respectively.
Both sets exhibit asymptotic limiting behavior to values below
a doubling of the initial bias voltage, the general figure of merit
Fig. 14. Bias voltage drift at 85 C, 10 Gb/s, x-cut, Ti waveguides [19].
used for the fail criteria in long-term aging studies. Fig. 12
shows the drift in bias point (measured in degrees of phase)
for x-cut 2.5-Gb/s APE waveguide devices also at 85 and 100 by phase drifts well beyond the values seen with these bias-free
C. These are bias-free device designs, where the operating designs.
point at 0 V bias has been set to quadrature through specialized Fig. 13 shows data from the literature for z-cut 10-Gb/s
manufacturing processes, so bias voltage is, by definition, titanium waveguide devices at various temperatures. These data
zero. Studies measuring the extinction ratio of pseudorandom also predict adequate reliability. Fig. 14 shows the drift of x-cut
bitstreams demonstrate that system performance is not affected 10-Gb/s titanium waveguide devices at 85 C. The significant
80 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

REFERENCES

[1] F. Heismann, S. K. Korotky, and J. J. Veselka, “Lithium niobate inte-

grated optics: Selected contemporary devices and system applications,”
in Optical Fiber Telecommunications III B, I. P. Kaminow and T. L.
Koch, Eds. New York: Academic, 1997, pp. 377–462.
[2] E. J. Murphy, “Photonic switching,” in Optical Fiber Telecommunica-
tions III B, I. P. Kaminow and T. L. Koch, Eds. New York: Academic,
1997, pp. 463–501.
[3] E. J. Murphy, Ed., Integrated Optical Circuits and Components: Design
and Applications. New York: Marcel Dekker, Aug. 1999.
[4] S. K. Korotky and R. C. Alferness, “Time and frequency domain re-
Fig. 15. Change in insertion loss with time, 85 C, 2.5 Gb/s, x-cut, Ti sponse of directional coupler travelling wave optical modulators,” J.
modulators [19]. Lightwave Technol., vol. LT-1, pp. 244–251, 1983.
[5] S. K. Korotky, “Three space representations of phase mismatch
switching in coupled two-state optical switching,” IEEE J. Quantum
Electron., vol. QE-22, pp. 952–958, 1986.
[6] R. V. Schmidt and I. P. Kaminow, “Metal-diffused optical waveguides
in LiNbO3,” Appl. Phys. Lett., vol. 25, pp. 458–460, 1974.
characteristic of Fig. 14 is the trend toward 0 V rather than the
[7] P. G. Suchoski, T. K. Findakly, and J. Leonberger, “Stable low-loss
usual increasing voltage with time. This strongly suggests that proton-exchanged LiNbO3 waveguide devices with no electro-optic
drift will not be a failure mechanism. The final curve, shown degradation,” Opt. Lett., vol. 13, pp. 1050–1052, 1988.
in Fig. 15, is the change in insertion loss for x-cut 2.5-Gb/s [8] J. L. Jackel, C. E. Rice, and J. Veselka, “Proton exchange for high-index
waveguides in LiNbO3,” Appl. Phys. Lett., vol. 41, pp. 607–608, 1982.
titanium waveguide devices at 85 C. As can be seen, loss [9] A. Yi-Yan, “Index instabilities in proton-exchanged LiNbO3 waveg-
changes are minimal as test time approaches 10 000 h. This uides,” Appl. Phys. Lett., vol. 42, pp. 633–635, 1983.
loss stability is a direct indicator of the reliability of the fiber [10] A. Yi-Yan, J. Primot, J. Burgeat, and R. Guglielmi, “Proton-exchanged
LiNbO3 waveguides: An X-ray analysis,” in Proc. 2nd ECIO, Florence,
to modulator interface and a strong indicator of the optical Italy, Oct. 1983, pp. 17–18.
circuit robustness. [11] R. A. Becker, “Comparison of guided-wave interferometric modulators
There are many other reliability and qualification tests fabricated on LiNbO3 via Ti indiffusion and proton exchange,” Appl.
Phys. Lett., vol. 43, pp. 131–133, 1983.
in addition to long-term aging studies. Telecordia (formerly [12] K. M. Kissa, P. G. Suchoski, and D. K. Lewis, “Accelerated aging of
Bellcore) TR-NWT-000 468 “Reliability assurance practices annealed proton-exchanged waveguides,” J. Lightwave Technol., vol. 13,
for optoelectronic devices in interoffice applications” [56] is pp. 1521–1529, 1995.
[13] D. K. Lewis, K. M. Kissa, and P. G. Suchoski, “Long lifetime projec-
the industry standard; however, it is about to be updated with tions for annealed proton-exchanged waveguide devices,” IEEE Photon.
GR-468-CORE “Generic reliability assurance requirements for Technol. Lett., vol. 7, pp. 1318–1320, 1995.
optoelectronic devices used in telecommunications equipment” [14] K. Noguchi, O. Mitomi, H. Miyazawa, and S. Seki, “A broadband
Ti : LiNbO3 optical modulator with a ridge structure,” J. Lightwave
[57]. This updated version includes, for the first time, external Technol., vol. 13, pp. 1164–1168, 1995.
modulators and consists of 13 different tests grouped into [15] F. Laurell, J. Webjorn, G. Arvidsson, and J. Holmberg, “Wet etching of
mechanical (mechanical shock, vibration, thermal shock, proton-exchanged lithium niobate - A novel processing technique,” J.
Lightwave Technol., vol. 10, pp. 1606–1609, 1992.
solderability, and fiber pull), endurance (accelerated aging, [16] W. L. Chen, R. S. Chen, J. H. Lee, and W. S. Wang, “Lithium nio-
high temperature storage, low temperature storage, temperature bate ridge waveguides by nickel diffusion and proton-exchanged wet
cycling, damp heat, and cyclic moisture resistance), and special etching,” IEEE Photon. Technol. Lett., vol. 7, pp. 1318–1320, 1995.
[17] K. Noguchi, O. Mitomi, K. Kawano, and M. Yanagibashi, “Highly effi-
(internal moisture and electrostatic discharge threshold) cate- cient 40-GHz bandwidth Ti : LiNbO3 optical modulator employing ridge
gories. It also distinguishes between hermetic and nonhermetic structure,” IEEE Photon. Technol. Lett., vol. 5, pp. 52–54, 1993.
components as many of the device types in commercial use [18] R. S. Moyer, R. Grencavich, F. F. Judd, R. C. Kershner, W. J. Minford,
and R. W. Smith, “Design and qualification of hermetically packaged
are not hermetic. Although passing a particular test does not lithium niobate optical modulator,” IEEE Trans. Comp., Packag., Man-
quantify reliability, it does take into account the extensive ufact. Technol. B, vol. 21, pp. 130– 135, 1998.
experience of Telecordia as to what is important to reduce the [19] D. Maack , “Reliability of lithium niobate Mach- Zehnder modulators
for digital optical fiber telecommunication systems,” in Proc. SPIE Crit-
risk of field failure problems. The data shown in Figs. 11, 12, ical Reviews: Reliability of Optical Fibers and Optical Fiber Systems,
and 14 are from qualification programs [19] in accordance with Boston, MA, Sept. 1999, pp. 197–230.
the Telecordia requirements. [20] A. O’Donnell, “Packaging and reliability of active optical components,”
in Proc. 7th Eur. Conf. Int. Opt. (ECIO ’95), 1995, p. 585.
[21] R. C. Alferness, “Waveguide electro-optic modulators,” IEEE Trans. Mi-
VIII. SUMMARY crowave Theory Tech., vol. MTT-30, pp. 1121–1137, Aug. 1982.
[22] R. V. Schmidt and R. C. Alferness, “Directional coupler switches, mod-
LiNbO3 modulators have found widespread use in fiber-optic ulators, and filters using alternating techniques,” IEEE Trans. Circuits
communication systems, including both chirped and zero-chirp Syst., vol. CAS-26, pp. 1099–1108, 1979.
[23] Y. Silberberg, P. Perlmutter, and J. E. Baran, “Digital optical switch,”
NRZ and RZ digital transmission formats. In addition, these Appl. Phys. Lett., vol. 51, pp. 1230–1232, 1987.
integrated-optic devices have proven to be extremely reliable, [24] W. K. Burns, A. B. Lee, and A. F. Milton, “Active branching waveguide
and device-manufacturing technology has progressed to enable modulator,” Appl. Phys. Lett., vol. 29, pp. 790–792, 1976.
[25] W. K. Burns, “Shaping the digital switch,” IEEE Photon. Technol. Lett.,
extensive deployment within digital and analog communication vol. 4, pp. 861–863, 1992.
systems. As the complexity of fiber-optic transmission systems [26] H. Okayama and M. Kawahara, “Prototype 32x32 optical switch ma-
and networks continues to grow, the capabilities and benefits trix,” Electron. Lett., vol. 30, pp. 1128–1129, 1994.
[27] T. O. Murphy, E. J. Murphy, and R. W. Irvin, “An 8 x 8 Ti : LiNbO3
of LiNbO3 modulators should continue to develop, thereby polarization independent photonic switch,” in Proc. Photonics in
providing new solutions and enabling new architectures. Switching/ECOC, Florence, Italy, 1994 , pp. 174–176.
WOOTEN et al.: REVIEW OF LITHIUM NIOBATE MODULATORS FOR FIBER-OPTIC COMMUNICATIONS 81

[28] K. Noguchi, O. Mitomi, and H. Miyazawa, “Millimeter-wave [54] C. Cox, E. Ackerman, R. Helkey, and G. Betts, “Techniques and per-
Ti : LiNbO3 optical modulators,” J. Lightwave Technol., vol. 16, pp. formance of intensity-modulation direct-detection analog optical links,”
615–619, 1998. IEEE Trans. Microwave Theory Tech., vol. 45, pp. 1375–1383, Aug.
[29] H. A. Wheeler, “Formulas for the skin effect,” Proc. IRE, vol. 30, pp. 1997.
412–424, 1942. [55] W. Minford, “The taming of LiNbO3,” in FIO III, St. Martin Vesubie,
[30] S.-J. Chang, C.-L. Tsai, Y.-B. Lin, J.-F. Liu, and W.-S. Wang, “Improved France, slide presentation.
electro-optic modulator with ridge structure in x-cut LiNbO3,” J. Light- [56] “Reliability assurance practices for optoelectronic devices in interoffice
wave Technol., vol. 17, pp. 843–847, 1999. applications,” Bellcore, TR-NWT-000 468, Dec. 1991.
[31] D. W. Dolfi and T. R. Ranganath, “50 GHz velocity-matched, broad [57] “Generic reliability assurance requirements for optoelectronic devices
wavelength LiNbO3 modulator with multimode active section,” in Proc. used in telecommunications equipment,” Bellcore, GR-468-CORE,
Integrated Photonics Research Conf., 1992, post-deadline paper PD-2. Dec. 1998.
[32] R. C. Alferness, S. K. Korotky, and E. A. J. Marcatilli, “Velocity
matching techniques for integrated optic traveling wave switch/modu-
lators,” J. Lightwave Technol., vol. 20, pp. 301–309, 1984.
[33] P. Hallemeier, K. Kissa, G. McBrien, and T. Horton , “Next generation Ed L. Wooten (M’92) received the B.S. degree in electrical engineering from
10Gb/s lithium niobate modulator components for RZ based transmis- the Massachusetts Institute of Technology, Cambridge, in 1982 and the M.S.
sion techniques,” in Proc. National Fiberoptic Engineers Conf., Sept. and Ph.D. degrees in applied physics from the University of California at San
1999. Diego, La Jolla, CA, in 1991.
[34] K. Imai, T. Tsuritani, N. Takeda, K. Tanaka, N. Edagawa, and M. Suzui, His graduate work concerned the modeling and fabrication of guided wave
“500 Gb/s (50x10 Gb/s) WDM transmission over 4000km using broad- electrooptic modulators in lithium niobate. From 1993 to 1997, he worked at
band EDFA’s and low dispersion slope fiber,” in Tech. Dig. Optical Fiber Tetra Tech Data Systems, Carlsbad, CA, on the development of a rapidly tunable
Communication Conf., Feb. 1999, post-deadline paper PD5. wavelength filter, fused fiber optic couplers, and a secure optical homodyne
[35] G. Gopalakrishnan, W. K. Burns, R. W. McElhanon, C. H. Bulmer, and communication system. In 1997, he joined the staff at JDS Uniphase, where he
A. S. Greenblatt, “Performance and modeling of broadband LiNbO3 is responsible for the design of custom LiNbO3 modulators and new product
traveling wave optical intensity modulators,” J. Lightwave Technol., vol. development.
12, pp. 1807–1819, 1994.
[36] S. K. Korotky and J. J. Veselka, “An RC network analysis of long
term Ti :LiNbO3 bias stability,” J. Lightwave Technol., vol. 14, pp.
2687–2697, 1996. Karl M. Kissa (S’86–M’88) received the B.S. degree from Duke University,
[37] N. S. Bergano, C. R. Davidson, C. J. Chen, B. Pederson, M. A. Mills, Durham, NC, in 1982 and the Ph.D. degree from University of Delaware,
N. Ramanujam, H. D. Kidorf, A. B. Puc, M. D. Levonas, and H. Ab- Newark, in 1989.
delkader, “640 Gb/s transmission of sixty-four 10 Gb/s WDM channels From 1989 to 1994, he was a Member of Technical Staff at Charles Stark
over 7200km with 0.33 (bits/s)/Hz spectral efficiency,” in Tech. Dig. Op- Draper Laboratory, Cambridge, MA, where he designed integrated optical (IO)
tical Fiber Communication Conf., Feb. 1999, post-deadline paper PD2. lithium niobate devices for fiber-optic gyroscopes. In 1994, he joined JDS
[38] W. K. Burns, “Linearized optical modulator with fifth order correction,” Uniphase, where he models and designs lithium niobate devices. He was also
J. Lightwave Technol., vol. 13, pp. 1724–1727, 1995. involved in the development of acousto-optic tunable filters.
[39] G. E. Betts, “Microwave analog optical links using suboctave linearized
modulators,” IEEE Photon. Technol. Lett., vol. 8, pp. 1273–1275, 1996.
[40] M. Izutsu, S. Shikama, and T. Sueta, “Integrated optical SSB modu-
lator/frequency shifter,” IEEE J. Quantum Electron., vol. QE-17, pp. Alfredo Yi-Yan received the B.Sc. and Ph.D. degrees in electronics and elec-
2225–2227, 1981. trical engineering from the University of Glasgow, Scotland, U.K., in 1974 and
[41] R. Ramaswami and K. Sivarajan, Optical Networks: A Practical Per- 1978, respectively.
spective. Los Altos, CA: Morgan Kaufmann, 1998. From 1978 to 1981, he was a SERC Research Assistant at the Electronics
[42] G. P. Agrawal, Non-Linear Fiber Optics, 2nd ed. New York: Aca- and Electrical Engineering Department, University of Glasgow, working on pe-
demic, 1995. riodic waveguide filters for integrated optics. In 1981, he joined CNET-Labo-
[43] F. M. Madani and K. Kikuchi, “Design of long-distance WDM disper- ratoire de Bagneux, Paris, France, and was engaged in work on lithium niobate
sion managed transmission system,” J. Lightwave Technol., vol. 17, pp. devices, as well as active and passive periodic waveguides in semiconductors. In
1326–1335, 1999. 1987, he left CNET to join Bellcore as a Member of Technical Staff, a position
[44] M. I. Hayee and A. E. Wilner, “NRZ versus RZ in 10-40 Gb/s dis- that he held until 1992. At Bellcore, his research interest focused on advanced
persion-managed WDM transmission systems,” IEEE Photon. Technol. optical components for fiber-optics communications systems and hybrid inte-
Lett., vol. 11, pp. 991–993, 1999. gration techniques. He has been with JDS Uniphase since 1997, where he is
[45] E. J. Murphy, “Photonic switching,” in Optical Fiber Telecommunica- currently engaged in process-development activities.
tions IIIB, I. P. Kaminow and T. L. Koch, Eds. New York: Academic,
1997, ch. 10, pp. 463–501.
[46] M. G. Taylor and S. J. Penticost, “Improvement in performance of long
haul EDFA using high frequency polarization modulation,” Electron. Edmond J. Murphy (SM’94) received the B.S. degree in chemistry from
Lett., vol. 30, no. 10, pp. 805–806, 1994. Boston College, Boston, MA, in 1976 and the Ph.D. degree in chemical physics
[47] R. Plastow, “Tunable lasers enable new optical networks to meet from Massachusetts Institute of Technology, Cambridge, in 1980.
changing demands,” Lightwave, pp. 74–79, Dec. 1998. From 1980 to 1999, he was a Distinguished Member of Technical Staff and
[48] W. Way, Broadband Hybrid Fiber/Coax Access System Technolo- Technical Manager at Lucent Technologies. At Lucent, he was involved with
gies. New York: Academic, 1999, ch. 7. work on optical amplifiers, lithium niobate devices, and optical data links. He
[49] G. Wilson, “Predistortion techniques for linearization of external mod- is currently Director of Technology for JDS Uniphase’s Electro-optic Product
ulators,” in Dig. IEEE/LEOS Summer Topical Meeting, San Diego, CA, Division in Bloomfield, CT. He holds 12 patents and is the author of more than
July 26–30, 1999, paper FB2.1., pp. 26–30. 50 technical publications.
[50] H. Skeie, “An optically linearized modulator for CATV applications,” Dr. Murphy is a Fellow of the Optical Society.
in Proc. SPIE, vol. 2291, 1994, pp. 227–238.
[51] W. Burns, M. Howerton, R. Moeller, R. McElhanon, and S. Greenblatt,
“Low drive voltage, 40 GHz LiNbO3 modulators,” in OFC ’99, San
Diego, CA, Feb. 1999, paper ThT1. Donald A. Lafaw received the B.S. degree (with high honors) and the M.S.
[52] K. Noguchi, O. Mitomi, and H. Miyazawa, “Low voltage and broadband degree in electrical engineering from the University of Illinois, Urbana, in 1983
Ti : LiNbO3 optical modulator operating in the millimeter wavelength and 1984, respectively.
region,” in Tech. Dig. Conf. Optical Fiber Communications, San Fran- From 1984 until 1997, he held various technical staff positions at the Rome
cisco, CA, Jan. 1996, paper ThB2, pp. 205–206. Laboratory Air Force Photonics Center, Lawrence Livermore National Labora-
[53] D. Lafaw, R. Logan, and F. Leonberger, “Multi-GHz components: An tory, National Security Agency, and the Laboratory for Physical Sciences at the
industrial perspective,” presented at the International Topical Meeting University of Maryland, College Park. He is currently a Business-Unit Manager
on Microwave Photonics, MWP ’99, Melbourne, Australia, Nov. 1999. for Advanced Digital and RF Products at JDS Uniphase in Bloomfield, CT.
82 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 1, JANUARY/FEBRUARY 2000

Peter F. Hallemeier (S’93–M’97) received the B.S.E.E. degree from North- Daniel J. Fritz received the B.S. degree in physics from the University of Hart-
eastern University, Boston, MA, and the M.S.E.E. degree from the University ford, West Hartford, CT.
of Rochester, Rochester, NY, with concentrations in electromagnetics, RF engi- From 1982 to 1995, he was a Member of Technical Staff at United Technolo-
neering, and optoelectronics. gies Research Center in East Hartford, CT. Since 1995, he has been employed
His research and design work has included high-temperature superconductor at JDS Uniphase in Bloomfield, CT, and is currently a Business Unit Manager
devices, liquid crystal devices, lithium niobate devices, and fiber systems. In for 2.5-Gb/s lithium niobate modulators.
1998, he joined JDS Uniphase in Bloomfield, CT and is currently a Business
Unit Manager.

Gregory J. McBrien (M’94) received the B.S.E.E. and M.S.E.E. degrees from
Rensselaer Polytechnic Institute, Troy, NY, in 1979 and 1997, respectively.
David Maack (M’97) received the B.S. degree in physics and in nuclear sci- He joined the advanced guidance and control electronics group at Hamilton
ence from Lowell Technological Institute, Lowell, MA, in 1969 and the M.B.A. Standard Division of United Technologies Corp. from 1979 to 1990, where he
degree from the University of New Haven, New Haven, CT, in 1983. worked on a number of fiber-optic sensor and data applications. He joined the
His career in fiber optics spans 25 years. He has held various developmental, United Technologies Research Center in 1990, as part of the Photonics Initiative
engineering, manufacturing, and management positions. He is currently Man- Group, and continued with what is now JDS Uniphase, working on a number
ager of Reliability Engineering at JDS Uniphase in Bloomfield, CT. His current of analog and digital development products involving lithium niobate. He has
interests include failure analysis and reliability predictions for optical compo- 13 patents in the areas of sensors, signal processing, and fiber-optic analog and
nents. digital technology areas.

Daniel V. Attanasio received the B.S. degree in physics from the State Univer- Donald E. Bossi (S’87–M’90) received the S.B., S.M., and Ph.D. degrees from
sity of New York at Albany in 1989 and the M.S. degree from the Institute of the Massachusetts Institute of Technology, Cambridge, MA, in 1985, 1986, and
Optics at the University of Rochester, Rochester, NY, in 1991. 1990, respectively, all in electrical engineering.
He was an Optical Engineer with 3M Corp. from 1990 to 1997, performing re- From 1990 to 1993, he worked at United Technologies Research Center in
search on LiNbO3 polarization-independent integrated-optical switches and po- East Hartford, CT, on the development of GaAs integrated optoelectronic cir-
larization-maintaining fiber-optic connectors. In 1997, he joined JDS Uniphase cuits. In 1994, he joined JDS Uniphase in Bloomfield, CT, where he is presently
in Bloomfield, CT, where he is currently a Product Development Project Leader. a Vice President of Technology and Development.

View publication stats