Appl Phys B (2009) 95: 421–428

DOI 10.1007/s00340-009-3501-4

Photorefractive effect in LiNbO3 -based integrated-optical circuits

at wavelengths of third telecom window
S.M. Kostritskii

Received: 24 November 2008 / Revised version: 4 March 2009 / Published online: 2 April 2009
© Springer-Verlag 2009

Abstract Original results on investigation of the photore- Many detailed studies were reported for straight wave-
fractive effect in straight channels and integrated-optical cir- guides, IO circuits and some devices operating in the visible
cuits such as a directional coupler, Y-splitter and Mach– and near-IR regions of the spectrum [1–4], but only two sys-
Zehnder interferometer, exploiting titanium-indiffused and tematic studies were performed at 1.3 μm [5, 6] and no such
proton-exchanged LiNbO3 waveguides, are presented. It study was reported yet for PRD in IO circuits at wavelengths
has been found that the photorefractive damage is non- around 1.5 μm, while some particular features were reported
negligible for IR radiation with wavelengths near 1.5 μm in only for the latter wavelength range [7–9]. However, recent
all circuits studied. The new methods for accurate evaluation activities in guided-wave electro-optic and nonlinear-optic
of small extents of photorefractive effect are proposed. devices are directed towards optical communication appli-
cations and concentrated in the wavelength region from 1.46
PACS 42.82.Bq · 42.65.Hw · 42.82.-m to 1.63 μm, where the photorefractive effect generally de-
creases and, thus, it has been neglected without any exper-
imental verification at design of all the LiNbO3 integrated-
1 Introduction optical devices for telecom and other applications. It may
be regarded as the right approach for dominating part of
The photorefractive damage (PRD) is a well-known classi- modern devices operating at low optical intensities. How-
cal problem for most electro-optic and nonlinear optic ap- ever, the advanced devices are operating with the powerful
plications of LiNbO3 crystals: Optically induced change of IR radiation (>100 mW) [9–11]. At the same time, even a
refractive index creates difficulties to produce stable devices very small refractive index variation, occurring at low in-
in this material for visible and near-IR ranges. This problem put power, is important to consider, if the circuit is used
is most dramatic in case of integrated-optical (IO) devices, in a phase-sensitive system such as a fiber optical gyro-
utilizing channel waveguides, due to the following specific scope [12]. Therefore, the IO LiNbO3 circuits must be de-
features: (1) Light beams are confined in two dimensions to signed so that effects of the photorefractivity are adequately
region only a few micrometers in size, and very high opti- compensated and knowledge of photorefractive properties is
cal intensities are thus obtained even at small input power; required in developing advanced devices.
(2) The inhibition of diffraction leads to effects, which do
not occur in bulk samples; (3) Waveguide fabrication means
heavy doping, as industrial IO LiNbO3 devices are fabri- 2 Sample fabrication techniques
cated only by the two techniques: Ti-indiffusion or proton
exchange. A series of straight channel waveguides, directional cou-
plers and Y-splitters, utilizing the different geometries of
Y-junction branching, were delineated in optical grade X-cut
S.M. Kostritskii () LiNbO3 substrates, using standard photolithographic tech-
MPTE Department, Moscow Institute of Electronic Technology,
124498 Moscow, Zelenograd, Russia
nique. The channel width W for Y-splitters and straight
e-mail: [email protected] waveguides was varied in the range from 4.6 to 10 μm.
422 S.M. Kostritskii

The one part of substrates was proton-exchanged at 175°C

for 50–70 min in pure benzoic acid; after that, substrates
were annealed at 360°C for 5.85–6.5 hours. In the second
part of substrates, an array of waveguide circuits was fab-
ricated by indiffusion of an array of Ti strips. The 100-nm-
thick strips were indiffused for 18–19.5 hours at the tem-
perature of 1000°C. To suppress Li2 O outdiffusion during
this process, the substrate was placed in Pt box along with
LiNO3 powder while dry Ar/O2 (1.2/1) gas was circulated
through the tube at 20 cm3 /min. During fast cooling-down
stage after Ti-indiffusion, the dry O2 was circulated through
the tube at 30 cm3 /min. Fig. 1 Steady-state output power versus input power for the
The IO chips of a symmetrical Mach–Zehnder inter- straight-channel Ti-indiffused LiNbO3 (1) and APE:LiNbO3 (2)
ferometer with 6-μm-wide arms and two Y-junctions of waveguides. The width of waveguides is 8 μm (1) and 6 μm (2)
constant branching angle (2.6◦ ) were fabricated by an-
nealed proton exchange (APE) or Ti-indiffusion technique ricated by different techniques, an effort was first made
in X-cut LN substrates at conditions mentioned in the previ- to examine optical transmission through straight channel
ous paragraph. The electrode structure consisted of a copla- waveguides under comparable conditions. To provide high
nar waveguide with three electrodes. The width of the cen- input intensities needed for measurements, a fiber Raman
tral electrode and the inter-electrode gap were chosen to be laser “IRE-Polus Group, FRL-1480-600” at 1.48 μm was
18 μm to arrive at a reasonable compromise between RF used. Light from the laser was coupled into the waveguides
ohmic loss in the electrode and the impedance matching using a single-mode fiber. Waveguide outputs were mea-
of the electrode structure to an external microwave source. sured using a 20× lens, pinhole, and infrared detector
Microwave bends and tapers were incorporated to provide with calibrated power meter. Output power versus input
transition from the electro-optic interaction area to the in- power was then plotted to check for photorefractive dam-
put/output ports. The total interaction length L was cho- age. Figure 1 shows the results of photorefractive measure-
sen to be 10 mm. The necessary electrode pattern was pho- ments performed on straight channel LiNbO3 waveguides.
tolithographically delineated in uniform Au-Cr seed layer The Ti-indiffused LiNbO3 waveguides show marked de-
that was deposited over the substrate. gree of saturation even at the moderate input power level.
The preliminary study of test samples of planar wave-
The significant attenuation of output power relative to in-
guides gave us the possibility to determine roughly a range
put power illustrates the high photoinduced refractive index
of fabrication parameters, where formation of a low-loss
change in the both types of LiNbO3 waveguides, especially
single mode channel waveguide, operating within a certain
above 100 mW.
wavelength region, is expected. A target wavelength of our
It has been found that the output power behavior de-
optimization procedure was ranged from 1520 to 1575 nm.
pends dramatically on the initial mode confinement in a
Final optimization of the fabrication parameters was made
channel guide, i.e., confinement observed before the start
according to data on direct measurements of the mode shape
of PRD. For example, a single-mode waveguide becomes
and insertion losses for each type of waveguides and IO cir-
more sensitive to photorefractive damage when it is closer
cuits.
to either fundamental-mode cut-off or double-mode regime.
Most of the IO chips were fabricated by polishing their
Thus, the measurement of output power attenuation cannot
ends at a sufficient slant angle (10◦ ) to avoid Fresnel back-
be used even for qualitative comparative study of a pho-
ward reflection from air–waveguide interfaces and, hence,
torefraction magnitude in the different waveguides, as the
to escape a parasitic Fabry–Perot interferometer. However,
same light-induced refractive index change may cause the
some IO chips with straight-channel waveguides have per-
different attenuation of output power, depending on initial
pendicular air–waveguide interfaces at the both optical-
light confinement. By the way, the qualitative studies of in-
grade polished lateral edges.
put power dependence of steady-state PRD and photorefrac-
tive response kinetics in a certain waveguide are possible
3 Experimental results and discussion to make, using these data. For example, the data shown in
Fig. 1 allow for conclusion that steady-state PRD depends
3.1 Experiments with straight channels: modal power on input power and this dependence is far from saturation
cut-off and phase retardation even at the largest input power used in both waveguides.
However, these measurements are incapable of generating a
To access the photorefractive effect at wavelengths of third quantitative measure of the steady-state photorefractive ef-
telecommunication window in the various waveguides fab- fect.
Photorefractive effect in LiNbO3 -based integrated-optical circuits at wavelengths of third telecom window 423

Fig. 2 Set-up for photorefractive effect measurement in waveguide

FPI (1), consisting of a straight channel LiNbO3 waveguide.
FRL—fiber Raman laser, TLS—tunable laser source HP 8168 (wave-
length range from 1525 to 1575 nm), S—switch, FS1:1—fiber splitter
with splitting ratio 1:1, SMF—single-mode fiber, OSA—optical spec- Fig. 3 Output power versus exposure time measured in FPI arrange-
tra analyzer ment with 6-μm wide Ti-indiffused LiNbO3 waveguide at 490-mW in-
put power. The relatively high insertion losses (>7 dB) observed even
at initial stage of this experiment are caused by an application of a
In order to quantify the extent of PRD, the refrac- photodiode (instead of OSA) with effective aperture limited by 4-mm
tive index changes were measured in the straight chan- pinhole
nel waveguides using a Fabry–Perot interferometer (FPI)
arrangement. The polished end faces of the waveguide form fact, the recorded temporal variations of P (t) were observed
the mirrors for the FPI. This experiment was carried out by to start at a fast rate but slow down with time as n∗ (t)
coupling the 1.48-μm irradiation of a powerful fiber Raman approaches saturation, Fig. 3. By counting the number of
laser (FRL) in a LiNbO3 waveguide and monitoring output fringe shifts in the recorded output, the resultant maximum
temporal power variation, Fig. 2. Note that in this experi- change in refractive index n∗ has been estimated to be
ment a closed-aperture photodiode was used to measure the about −8 × 10−5 and −9.6 × 10−5 in APE LiNbO3 and
output power instead of OSA, a high numerical aperture mi- Ti:LiNbO3 waveguides, respectively, at 490-mW input.
croscope objective 20× was applied, instead of output fiber, Since there is no strong absorption of the pump light used
to collect the output radiation. having wavelengths of around 1.48 μm, thermal effects due
The change of transmitted beam power, P (t), depends to absorption of the pump light are considered to be small
on the effective refractive index of a guided mode, n∗ (t), compared to the PRD. This assumption is supported by our
and wavelength, λ, as: data on the dark decay of the light-induced effects, in which
  it was found that a recovery to the initial state occurs for
P (t) = (Pmax − Pmin ) sin2 2πn∗ (t)L/λ (1) a period of the order of ten hours. Note that the specific
time of the dark decay of thermal effects is of the order of
where Pmax and Pmin are the maximum and minimum output minutes [1]. It has been therefore concluded that although
power at a certain stage of PRD, L is the waveguide length, thermal effects may give some contribution, the observed
and (Pmax − Pmin ) depends on propagation loss [9]. FPI fringe shifts are due to, dominantly, PRD in the LiNbO3
The irradiation excites the electrons and leads to the waveguide.
charge separation and development of a temporal space According to the theoretical and experimental findings
charge field, Esc (t), that induces index change, n∗ (t), for LiNbO3 waveguide FPI [13], the significant decrease of
through the linear electro-optic effect [1]: the modulation depth Km = (Pm,max − Pm,min )/(Pm,max +
Pm,min ) with exposure, i.e. with a fringe shift number m, in-
n∗ (t) = −0.5n3 r33 × Esc (t) dicates the growth of propagation losses ε (extinction coef-
   
= − (σph /σ )Eapp + αGI /σ × 1 − e−t/τ (2) ficient) caused by light-induced scattering due to PRD with
specific kinetics (2).
where r33 is the electro-optic coefficient, G is Glass con- The light-induced increment ε of extinction coefficient
stant, α is optical absorption coefficient, αGI is photo- may be evaluated by ε = 4.34K/K0 L, where K0 is
voltaic current density, σph = αβph I is the photoconductiv- modulation contrast observed at initial stage of this exper-
ity, βph is specific photoconductivity parameter, τ = εr ε0 /σ iment. For example, experimental data on K and K0 for
is the build-up time constant, σ = σd + σph is conductivity, waveguide FPI shown in Fig. 3 give estimation of ε value
σd is the dark conductivity, I is light intensity into guide and of about 0.4 dB/cm, which is quite in satisfactory agreement
Eapp is external electric field. with our data on direct measurements of insertion optical
Both Esc (t) and n∗ (t) saturate eventually with time. losses (increment is about 0.7 dB/cm at 490 mW) obtained
The output power of transmitted beam should be modu- in experiments similar to the one shown in Fig. 1. Such coin-
lated periodically in time, since n∗ (t) = n∗ (0) + n(t). In cidence is observed in all the studied FPIs of both waveguide
424 S.M. Kostritskii

types. The larger magnitude of light-induced increase of the

insertion losses in comparison with the propagation losses is
observed only at high input power (≥100 mW) and it may
be related to the light-induced increase of a mode size at the
near-field output intensity profile. Such an increase of the
fundamental mode width was established to be very signif-
icant at propagation of laser irradiation of the visible range
in Ti:LiNbO3 waveguides [14].
The application of this method at low input power is a
rather difficult task as n∗ may be not sufficient to induce
even one fringe shift at a real guide length. To solve this Fig. 4 Spectral dependence of output power of the weak probe
problem, we have developed a new method for more ac- beam (input power was 30 μW) for the FPI fabricated on basis
curate FPI experiments. Experiments were carried out by of single-mode 8-μm-width straight-channel Ti-indiffused LiNbO3
coupling two optical beams: a probe beam from a tun- waveguide. FPI cavity length is 34.4 mm, channel width is 7 μm. The
spectral dependence shown is obtained after extra exposure via power-
able laser source (TLS) to monitor the intensity transmitted ful pump beam with input power of 15 mW for 3.5 hours. The latter
through the waveguides, and an irradiation beam from FRL exposure induces the decrease of wavelengths of minimum and maxi-
at 1.48 μm to induce the photorefractive damage. The po- mum output power for each interference fringe for ∼8 pm
larization of each beam was oriented to excite the TE mode
in the X-cut waveguides. A polarization maintaining fiber the experiments mentioned above (see Fig. 3), and small de-
splitter (FS) was used to co-launch the two beams simul- viation of a sequence of interference fringes from an exactly
taneously. The wavelength dependence of transmission for periodic law, Fig. 4. The latter may be related to the weak
probe beam from TLS, having the 1-pm step, was measured photorefractive intermode conversion TE0 → TM 0 , accord-
with aid of an optical spectrum analyzer HP 70951A. Dur- ing to the previous findings [1].
ing these spectral measurements, the FRL was disconnected Note that the accuracy of measurements of both the posi-
by switch (S) from an input port of FS and output power of tion λm m
p (0) and its shift λp is increasing with L. There-
the probe beam of TLS was kept below 50 μW to minimize fore, to determine the changes of small photoinduced re-
extra damage during spectral measurements. fractive index in LiNbO3 waveguides, we used the rela-
Our measurements before the start of photorefractive tively large samples with L > 2 cm and, thus, reached the
damage, i.e. before first launch of irradiation beam into high resolution in determination of n(t). According to the
waveguide, show that the spectral response of waveguide smallest wavelength step of 1 pm possible for our tunable
FPI presents a periodic smooth dependence of output power laser, the highest n resolution reached by this method is
on wavelength of input probe beam with a fixed sequence of about 1.3 × 10−6 . Thus, we have determined that the steady-
maxima and minima, specific for each waveguide. Impact state value n∗s (i.e., when dependence of n∗ on t is
of irradiation beam represents the shift of an interference
saturated) of −10−5 was reached in Ti-indiffused LiNbO3
spectrum on value λp , which is gradually increased by ex-
and APE:LiNbO3 waveguides with input power of 14 and
posure via the irradiation beam, Fig. 4. This spectral depen-
60 mW, respectively. This value of n∗ was determined as
dence of output power was mathematically fitted with high
critical level for performance degradation of most important
accuracy, using (1). It allows to determine a small optically
IO circuits [1, 7–9].
induced refractive index change n∗ with very high accu-
According to the commonly used band transport model
racy, if a resulting phase shift at any arbitrary fixed wave-
[1, 4, 11], the amount of n∗s should increase with the
length, λp , is smaller than π/2. In this case, the value of
guided power and reach a constant value at high power, de-
n∗ may be evaluated from the experimental data on the
pending on the ratio between photo- and dark-conductivity,
gradual shift λp of each interference fringe as:
and following the function n∗s = −[(aP )/(b + cP )],
 
n∗ (t) = n∗ (0) λm m
p /λp (0) (3) where P is the guided power and a, b and c are con-
stants, e.g., a = 9.465 × 10−7 mW−1 , b = 0.9913 and
where λm m
p (0) and λp are positions at start of PRD and c = 8.984 × 10−3 mW−1 in APE waveguides at excita-
shift, because of photorefraction for certain interference tion wavelength of 1.064 μm [11]. The saturation occurs at
fringe, m, in spectral dependence of output probe beam of the value a/c, when cP  b (i.e., at the guided power P
damage for certain interference fringe, m, in spectral depen- of order of several Watts) and it is estimated to be at
dence of output probe beam. n∗s ∼= −10−4 . Previous experimental findings [1–8, 11] re-
Besides, the PRD induced by pump beam is accompanied ported for wavelength range from 0.4 to 1.1 μm demonstrate
by reduction of the modulation contrast K of spectral depen- that the coefficient c related to the product of specific photo-
dence, similarly to temporal dependence of output power in conductivity and optical absorption sharply decreases with
Photorefractive effect in LiNbO3 -based integrated-optical circuits at wavelengths of third telecom window 425

excitation wavelength increase, while coefficient b related β  = β + φr is the difference between propagation
to the dark conductivity is, indeed, independent of wave- constants in photorefractive waveguide, φr is relative
length. Note that dark conductivity in the both Ti-indiffused phase retardation caused by photorefractive effect, β =
and proton-exchanged LiNbO3 is much higher than in a β0 − β1 , and lc = π/β is the critical coupling length
bulk LiNbO3 material [4]. Hence it might be expected that needed to couple all of the power from one guide to the other
a higher power level is required to be close to saturation in the absence of photorefractive effect, i.e., at n∗ (t) ≈ 0.
of n∗s (P ) dependence at 1.5-μm excitation, and maximal A critical parameter of active directional coupler, i.e.
P = 490 mW used by us is significantly smaller than this electro-optically controlled amplitude modulator (switch), is
level. the crosstalk of the switching in the crossed state, when all
of the light coupled into the second guide (P2 has the max-
3.2 Directional couplers imally possible value, and P1 → 0, Fig. 5). When noted in
decibels (dB), the crosstalk is:
The comparative study of the different 1.5-μm-band  
integrated-optical LiNbO3 circuits has showed that the di- CT = 10 × lg P1 /(P1 + P2 ) (5)
rectional couplers are most sensitive to the photoinduced
damage effects. This finding coincides with all the previous where P1 and P2 are the output powers of the initial guide
comparative studies performed at the different wavelengths and the coupled guide.
of visible and near-IR ranges [1, 8, 9]. The significant effect of the photoinduced index change
A passive directional coupler shown in Fig. 5 is designed upon crosstalk of a Ti:LiNbO3 directional coupler was ob-
so that each arm of the coupler is a single mode at the served previously [1, 5, 6] in the wide wavelength range
wavelength of interest, in this case ∼1.5 μm. Thus, the cou- from 0.633 to 1.3 μm. Thus, to achieve a crosstalk level
pler supports the symmetric and antisymmetric supermodes, of −30 dB at 1.3 μm in the 8-mm-long (7-μm-gap) active
which have different propagation constants β0 and β1 , re- Ti-indiffused directional coupler, the light intensity should
spectively. These passive couplers were observed to undergo be kept below 5 × 102 W/cm2 and −20 dB level is ob-
asynchronous, oscillatory power exchange, deviating from served at about 2 × 103 W/cm2 , which corresponds to ex-
its stable baseline signal at high input power. It has been tremely low input power of ≤150 and ≤600 μW, respec-
established that a fraction of the total energy oscillates be- tively [6]. Our experimental data show that the active direc-
tween the throughput P1 and crossover P2 channels, but it is tional couplers, operating at wavelengths near 1.5 μm, have
never totally transferred to the crossover channel. Amplitude a higher threshold of the photorefractive damage: a steady-
of this oscillation is gradually damped with time and some state crosstalk level of −20 dB is achieved at input power
steady-state stage is reached finally with an arbitrary value of about 10 and 35 mW in Ti-indiffused and APE LiNbO3
of the ratio P1 /P2 . Kinetics of the latter saturation process is couplers, respectively.
well described by the build-up time of photorefractive effect
(see (2)). Thus, it may be concluded, that effective or critical 3.3 Y-junction branching and Y-splitters
coupling length lc of the directional coupler is increased due
to photorefraction, as from coupled-mode theory [15, 16], Optical waveguides in lithium niobate are a very useful tech-
the output power ratio of this coupler being expressed as nique for building a variety of IO components. For exam-
ple, the key component of fiber optical gyroscope (FOG) is
P2 /P1 = tan2 (πL/2lc ) = tan2 (β  L/2 + θr ) the multifunction integrated optics chip (MIOC), which is
  an important part of the Sagnac interferometer [12]. As the
= tan2 πL/2lc + 2πLn∗ (t)/λ + θrc (4)
MIOC consists inherently an Y-junction branching, the pho-
where θrc represents the amount of residual coupling which torefractive effect can present the more dramatic problem,
occurs over the tapered region after the interaction length L, comparing to straight channel, if a change of refractive in-
dex is sufficient to induce even a small variation in the mode
coupling between two channel waveguides, representing the
arms of Y-junction near splitting point that can be regarded
as a directional coupler with short coupling length [15].
To fabricate a low-loss Y-splitter, we utilize a Y-junction
(section II) formed by three single-mode channel wave-
guides with width W = 6 μm (sections I and III), Fig. 6. The
Y-junction topology is described by the following equation:
Fig. 5 Illustration of a passive directional coupler including 7-μm
ye − ys
channels with 7-μm gap and interaction length L. Pin —input power, y(x) = ys + (x − xs ) (6)
P1 and P2 —output throughput and crossover powers xe − xs
426 S.M. Kostritskii
Fig. 6 Schematic illustration of
Y-junction splitter utilizing a
Y-junction (section II) formed
by three single-mode channel
waveguides (sections I and III)
with channel width W

A perfect Y-junction power divider should have

PTC = 0.5, and deviation of initial PTC (PTC0 observed
at t → 0, or at any t with a low input power ranged from 30
to 100 μW) from this value is caused by parasitic asym-
metry of Y-branching section due to technological imper-
fections and modes of interaction, as this branching section
may be regarded as analog of the directional coupler with
weighted coupling [15]; i.e. a normalized deviation para-
meter (0.5-PTC) is proportional to β/θ γ3 , where β is
accidental asymmetry of Y-junction caused by technolog-
ical imperfections, Θ is branching angle, γ3 is transverse
component of phase constant. Hence, our experimental data
Fig. 7 Dependence of the steady-state value of power trans-
fer coefficient (PTC) on input power Pin for Y-junction power allow for assumption that γ3 is altered by photorefractive
dividers (branching angle θ = 1.9◦ , W = 6 μm, see Fig. 6), effect.
utilizing the different types of channel LiNbO3 waveguides: In general, the following statement may be derived
1—proton-exchanged with PTC0 = 0.497; 2—Ti-indiffused with from experiments with the Y-branching power dividers
PTC0 = 0.495; 3—proton-exchanged with PTC0 = 0.488. Variation
of PTC0 values is caused mainly by imperfections of photolithography based on Ti-indiffused and APE LiNbO3 waveguides: a
process, e.g., the dividers #1 and 3 were fabricated at the same techno- smaller value of PTC0 is a higher susceptibility to PRD,
logical conditions. PTC0 is evaluated as a ratio between smaller output i.e. |∂(PTC)/∂Pin | is larger at the moderate input power.
power and the sum of both output powers measured at input power Experimental study of different power dividers with similar
≤2 mW
PTC0 shows that |∂(PTC)/∂Pin | is larger at application of
Ti-indiffused LiNbO3 waveguides in comparison with the
proton-exchanged ones. The attenuation of PTC has power-
where xe and ye have values of 11 and 0.16 mm, respec-
dependent kinetics specific for PRD (2) that is the most dan-
tively. The width of tapered subsection of Y-junction’s sec- gerous feature of a photorefractive power divider at applica-
tion II is 2W near the splitting point, having coordinates xs tion in phase-sensitive systems. Therefore, even a very small
and ys . Section II presents the analog of the directional cou- temporal variation of PTC (≥10−3 ), occurring at low input
pler with weighted coupling. power, may present dramatic problem for some particular
The power transfer coefficient (PTC) is established to applications.
be critical parameter PTC of the Y-junction branching, as Thus, consideration of the photorefractive effect at a care-
this parameter is most sensitive to PRD influence. PTC is ful design of the Y-junction branching is mandatory to ob-
evaluated experimentally as a ratio between smaller output tain the high performance MIOC. According to the theory
power and the sum of both output powers, i.e. PTC is di- of Y-junction [15, 16], to minimize a light-induced change
rectly related to a splitting ratio, which is the main parameter of PTC, the branching angle θ may be increased, but a
characterizing performance of the power divider. For exam- branching loss (αb ) will grow. Thus, θ = 1.9◦ is evaluated
ple, the marked attenuation of PTC, i.e. significant degrada- as an optimal value for the low input power (≤10 mW),
tion of power dividers’ performance, is observed in all the when αb = 0.6 dB. The value of branching angle θ near
1.5-μm-band Y-branching power dividers even at moderate 3.4◦ is theoretically determined as more suitable for high-
input power (50–250 mW), Fig. 7. This effect is observed power application, as Y-splitter should be more stable rel-
at the first time, i.e. when the new consequence of PRD is ative to influence of photorefractivity, i.e. PTC is expected
discovered. to be practically constant at input power of ≤120 mW, but
Photorefractive effect in LiNbO3 -based integrated-optical circuits at wavelengths of third telecom window 427

αb = 1.3 dB. Further increase of θ is out of practical interest, The most dramatic photoinduced change of MER was ob-
as it will induce sharp growth of αb . served in MZ interferometric Ti-indiffused LiNbO3 modu-
According to the theory of closed-loop FOG [12], the ra- lator with the initial MER = 25 dB, as steady-state value of
tio PTC/PTC0 for light-induced change of Y-splitter per- MER observed at Pin = 120 mW was 19 dB.
formance should be ≤0.02 at application of MIOC in a high- To develop theoretical explanation of input-power depen-
precision FOG (bias drift ≤0.01◦ /h). It means that output dent MER, we consider that as Mach–Zehnder interferome-
power of superluminescent diode, which is used as a light ter consists of two Y-junction splitters, MER should depend
source with the center wavelength of 1.52–1.56 μm, should on PTC of Y-junction splitters and asymmetry of interfer-
be limited by 13 and 55 mW in Ti-indiffused and proton- ometer arms δ [16, 17]:
exchanged MIOCs, respectively.    1/2 
Pmax /Pmin = 1 + δ + 1 − δ 2 1 − 2 /( + δ)2
3.4 Mach–Zehnder interferometers and interferometric (8)
modulators
where  = (1 − 2 × PTC). Note that we assume that both
Since IO Mach–Zehnder (MZ) interferometer is symmet- Y-junction splitters have the same PTC value.
ric two-arms circuit, it should be most stable against small Thus, our model explains the light-induced degrada-
and moderate extent of photorefractive damage, which was tion of MER as consequence of PTC change (see Fig. 7)
confirmed experimentally at 1.3 [7, 8] and 1.48 μm [9]. caused by photorefractive effect in Y-junction splitters con-
However, some undesirable consequence of photorefractive sisting of MZ interferometer. Moreover, there is a rather
effect (e.g., DC drift [8, 9]) becomes evident only at ap- good quantitative correlation between theoretical and exper-
plication of MZ interferometers in electro-optical intensity imental data. For example, (7), (8) give MER = 29 dB at
modulators. Therefore, our investigation was focused on the δ = 0.06 and PTC = 0.495, while MER = 19 dB is obtained
input-power dependence of switching capability of the (MZ) at PTC = 0.465 and δ = 0.06. Such a variation of PTC is in
interferometeric modulators fabricated in X-cut LiNbO3 . good accordance with data on light-induced changes of PTC
To characterize the switching capability, we measure the for Y-junction splitters, Fig. 7.
modulation extinction ratio (MER) at the steady state, i.e.
after long-term (order of few or tens hours) illumination:
4 Conclusion
MER = 10 × lg(Pmax /Pmin ) (7)
The photorefractive effect is found to be non-negligible in
where Pmax and Pmin are maximum and minimum values of the integrated-optical LiNbO3 circuits intended for practical
output power observed at variation of external DC-voltage applications at wavelengths of third telecommunication win-
applied on modulator electrodes. dow. The directional couplers, Y-branching power dividers
To measure Pmax and Pmin , the DC-voltage was gradually and MZ interferometers exhibit the light-induced degrada-
varied within the range from −34 to +34 V, inducing a cos2 - tion of important parameters even at a moderate input power,
like dependence of output optical power on applied voltage. where light-induced cut-off of guided light is not observed.
This dependence has three maximums and two minimums Besides, the influence of photorefractive effect on splitting
for the output power. ratio of Y-junction and modulation contrast of MZ interfer-
Our experimental study shows that MER depends on the ometric modulators have been observed at the first time.
input power Pin at moderate and high power levels in all our
MZ interferometric modulators. However, quantitative char-
acter of this dependence is very different even for MZ modu- References
lators fabricated at the same technological conditions. Such
a particularity has been attributed to the inherent asymmetry 1. V.E. Wood, P.J. Cressman, R.L. Holman, C.M. Veber, Photorefrac-
of Y-junction regions, appearing due to technological uncer- tive waveguides, in Topics Appl. Phys., vol. 61, ed. by P. Gunter,
tainty of photolithography and etching processes. In general, J.-P. Huignard (Springer, Berlin, 1989), pp. 45–100
2. R.A. Becker, R.C. Williamson, Appl. Phys. Lett. 47, 1024 (1985)
the following statement may be derived from experiments
3. R.A. Becker, Proc. SPIE 578, 12 (1985)
with various modulators: a smaller value of initial MER (ob- 4. T. Fujiwara, R. Srivastava, X. Cao, R.V. Ramaswamy, Opt. Lett.
served at t → 0, or at any t with a low input power ranging 18, 346 (1993)
from 50 to 100 μW) is a higher susceptibility to photore- 5. G.T. Harvey, G. Astfalk, A.Y. Feldblum, B. Kassahuh, J. Quantum
Electron. 22, 939 (1986)
fractive damage. This statement is valid in comparison of
6. G.T. Harvey, J. Lightwave Technol. 6, 872 (1988)
the same type (proton-exchanged or Ti-indiffused) LiNbO3 7. Y. Fujii, Y. Otsuka, A. Ikeda, IEICE Trans. Electron. E90-C, 1081
modulators with quite large initial MER values of ≥25 dB. (2007)
428 S.M. Kostritskii

8. D.B. Maring, S.M. Kostritskii, R.F. Tavlykaev, R.V. Ramaswamy, 13. R. Renger, W. Sohler, Appl. Phys. B 36, 143 (1985)
Integrated Photonics Research, OSA Technical digest, RTuK4-1, 14. J.C. Chon, W. Feng, A.R. Mickelson, Appl. Opt. 32, 7572–7590
281 (1999) (1993)
9. S.M. Kostritskii, in Technical Digest of the Workshop Lithium 15. H. Saaki, I. Anderson, IEEE J. Quantum Electron. QE-1, 883
Niobate from Material to Device, from Device to System, Metz, (1978)
France, May 23–25, 2005, pp. 59–64 (2005) 16. W.K. Burns, A.F. Milton, Waveguide transitions and junctions, in
10. N.A. Riza (ed.), Photonic Control Systems for Phased Array An- Guided-Wave Optoelectronics, ed. by T. Tamir (Springer, Berlin,
tennas. SPIE Milestone Series, MS 136 (1997) 1988), pp. 89–144
11. J.-P. Ruske, B. Zeitner, A. Tunnermann, A.S. Rasch, Electron. 17. H. Nishihara, M. Haruna, T. Suhara, Optical Integrated Circuits
Lett. 39, 1048 (2003) (McGraw-Hill, New York, 1985), pp. 36–44
12. H. Lefevre, The Fiber–Optic Gyroscope (Artech House, Norwood,
1993)