You are on page 1of 18

Home

Search

Collections

Journals

About

Contact us

My IOPscience

An introduction to methods of periodic poling for second-harmonic generation

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 1995 J. Phys. D: Appl. Phys. 28 1747 (http://iopscience.iop.org/0022-3727/28/9/001) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 164.125.41.50 The article was downloaded on 25/01/2013 at 05:13

Please note that terms and conditions apply.

J. Phys. D Appl. Phys. 28 (1995) 1747-1763. Printed in the UK

REVIEW ARTICLE

An introduction to methods of periodic poling for second-harmonic generation
M
Houe and P D Townsend
School of Mathematical and Physical Sciences, University of Sussex, Brighton BN1 9QH, UK Received 20 March 1995, in final form 16 June 1995 Abstract. Second-harmonic generation (SHG) can be produced by phase matching using the birefringence of nonlinear crystals via the modal dispersion in the case of optical waveguides. Such an approach limits the range of frequencies which can be doubled and also the choice of the nonlinear coefficients. One solution to both problems is to modify the crystal so as to have regions of periodic domain polarity. Whilst this approach does not allow a perfect phase match between the.fundamental and harmonic, it nevertheless can be entirely constructive throughout the interaction length of the material and is termed quasi-phase matching (QPM). Periodic modulation of the nonlinear coefficient along the direction of propagation can achieve conversion efficiencies up to 20 times greater than with previous methods. Candidates of interest for quasi-phase-matching are wide band gap inorganic crystals such as LiNbOs, LiTaOs and KTP, and also organic materials if they are transparent, stable against optical damage and have large nonlinear coefficients. To achieve QPM a variety of methods are being tried in order to invert domains periodically, either during the crystal growth phase, or subsequently by altering the lattice of the crystal. For inorganic ferroelectrics most effort has been concentrated on domain inversion in LiNb03 and LiTa03. Techniques have included application of pulsed electric fields, fields generated during electron bombardment, thermal pulsing or chemically driven movement of lithium. Many of the methods are semi-empirical in that the mechanisms by which the lattice re-structures are poorly understood. This review will therefore not only list the methods that are currently being used, but also comment on the underlying physical processes which allow, or prevent, the restructuring of the lattice and the domain walls, whilst preserving the non-centrosymmetric characteristics of the lattice. An understanding of mechanisms is valuable for related poling applications in other crystals and it is furlher noted that many amorpho-us systems, including glasses used for optical fibre communication, may be stimulated to show periodic structural changes although the usage precedes the knowledge of the mechanisms. The commercial applications and research possibilities for efficient SHG guarantee that this topic area will continue to be central to photonics for a considerable time.

1. Introduction
Photonics encompasses a rapidly expanding range of new materials, techniques and optical components including familiar examples such as solid state lasers, optical fibre communication, bar code laser scanners, information storage and compact disc players. The enormous commercial possibilities have resulted in a worldwide effort to develop higher performance optoelectronic devices, and, commensurate with this, there i s a drive to develop better nonlinear materials, to use existing ones with higher efficiency and to reduce costs. Since many of the applications would benefit from the availability of low-cost
0022-3727/95/091747+17519.50 0 1995 IOP Publishing Ltd

lasers which operate at the blue, or shorter wavelength, end of the spectrum, there is a significant effort to achieve this directly by production of semiconductor lasers, for example from U-VI compounds. Alternatively one might reach the same spectral region by frequency doubling from existing lasers (such as ID-V diode lasers). However, to make this commercially interesting, one fnrther needs high-efficiency nonlinear crystals, but development of such materials is a relatively long-term process and often expensive. Hence, there has been much attention devoted to frequency doubling with well-established materials such as lithium niobate. Frequency doubling in nonlinear materials has been extensively discussed in many excellent

1747

M Houe and P D Townsend

2.45

2.40

2.35
X

a , 2.30
C ._

._

9

2.25

t
a ,

( [ I

2.20

2.15

2.10
I

I

I

,

I

I

.

I

0.0

0.5

1.0

1.5

2.0

2 . 5

3 . 0

Wavelength

(microns)

Figure 1. Refractive indices of LiNbOa and a transmission curve for unpolarized light (Crystal Technology Inc).

Z(OPT1CS) A X I S

I

Rgure 2. Index ellipsoids of LiNbOsshowing the conditions needed for
1748

SHG,

Periodicity poling for second-harmonic generation books and articles [ I N ] but for completeness we will briefly review the essential features of second-harmonic generation (SHG) and examples for both single crystals and optical waveguides will be included. From these one can realize the material limitations in that there are short-wavelength limits and that for practical reasons one might not be able to exploit the highest valued nonlinear coefficients. Such problems are not new and as early as 1962 there was a proposal [9] that, rather than rely on perfect phase coherence between the driving frequency and the wavelets of the second harmonic, one could compromise and allow a steady phase slippage between them. If no further action were taken, then such phase differences would result in signal cancellation and an effective sample length equal to the coherence length Because this would not give efficient SHG, the suggestion was that, by modification of the crystal, the sign of the phase be inverted before cancellation commenced. This ensures that the overall conversion length is limited only by the length of the sample, which is crucial since the SHG conversion efficiency is proportional to the length squared. The method is known as quasi-phase matching (QPM) and it can be achieved if one can periodically invert the polarity of the crystal domains. With this objective, there have been many methods devised to grow, or alter, crystal domains with repetitive inversion, and examples of these ideas and mechanisms by which re-poling can be achieved are the subject of this progress report on periodic poling. Note also that, rather than meet the challenge of domain inversion, one may reach nearly the same condition by modifications which provide periodic alterations between nonlinear and linear material. Although the discussions are presented in terms of SHG, precisely the same advantages of QPM accrue to applications of difference frequency mixing, or other parametric interactions [lo].
2. Nonlinear optics in bulk materials

crystal orientation by inlroducing the nonlinear coefficients d.. - xij/2. to obtain

,

dii

4 2

d2i

d22

41

d 3 2

dis dzh dzs d26 d33 d34 d3s d36
dl3
did

d23

dl'l
(4)

The matrix can usually be simplified, since symmetry allows cancellation or equalization of certain coefficients. For instance, for one of the key optoelectronic materials Limo3,the matrix is reduced to

[px (b) 1

4

0
2

P y W ) - 20

0 d22

0
0

0 d33

41
0

-&I
0

(5)

There is considerable variation in the values cited for the coefficients and the spread is typically from 2.3 to 3.46.7 pm V-l for d22, 4.76 to 8.6-10 pm V-' for 4 1 and 30.8 to 32.2 pm V-l for d33 [S, 8,11,12]. The sign of the nonlinear coefficient determines the sense of the polarization and the overall efficiency of such harmonic generation is given [8] as r l = -pz, =22 PO

I
(EO)

3f2

w2d2l2sinz(Ak1/2) P, n3

(Ak1/2)

A

(6)

For intense fields the simple relation P = EOXE is no longer valid and several conhibutions have to be taken into account to work out the induced polarization P of the material. Effectively one must use the expanded form (1) and note that the susceptibility is a tensor. If the initial field is of the form E(r, t ) = ;[E(r)e'" P ( r , t ) = Eo[XiE(r, t ) = e ( r , t ) Pz(r,t)

+

+ x2E2(r. t) + x3E30, t ) + .I

+ P3(r,t ) +...

t .

+ cc] + cc].

(2)

where cc stands for complex conjugate, then, insofar as the second order terms are concerned, we have
Pz(r, t ) = &oX2E2(r,t) = EoXz[E2(r)e-'"
(3)

Therefore, if the term E ~ x is ~sufficiently E ~ large then the fundamental wave generates a second harmonic. The scale of the nonlinearities is controlled by the crystal structure and the polarization of the incoming light. We can quantify the SHG performance for a particular material or

where Pm and P z , are the fundamental and the secondharmonic powers respectively, d the nonlinear coefficient, 1 the interaction length, n3 = n i n z , the refractive index, k the wavevector of the fundamental in the medium and A the cross sectional area of the pump beam. Since the efficiency increases with the power density, it may be advantageous to focus the beam within the crystal. Focusing of the beam modifies the efficiency via an effective change in the value for the area which can also include compensation for a Gaussian beam profile. The efficiency expression emphasizes that ideally Ak = 0 which can only occur if 2k(o) = k(2w). Note that the interference terms sin2(Akl/2)/(Ak1/2)2 can be written as sinc2(Akl/2). Failure to minimize Ak means that there will be destructive interference between wavelets generated at different points in the crystal. There is thus aspatial coherence length given by , I I , = - =2 K (7) Ak 4(nz, - n o ) ' Reference to the dispersion data for a nonlinear crystal such as LiNb03 shows (figure 1) that one can use the natural birefringence to equate the n, and no indices for two polarizations of light, whose wavelengths differ by a factor of two, at a pump wavelength of 1100 nm. For
1749

M Houe and P D Townsend

v1

0 ._ -U
C -

a

1.54

-

a l U
0

I 1.52

-

I1S O
0.0

I
I

I

/ /

,
1.o

I

I

2.0

3.0

4.0

D e p t h (microns) Figure 3. Modal dispersion curves WO wavelengths as a function of waveguide depth. Note that the mode index crossings points allow phase matching. The example is calculated for 400 and 800 nm light in quartz. negatively birefringent crystals (no > ne) the use of a pure ordinary wave for the pump is termed type I phase matching. However, if the incident wave is inclined at an angle relative to the primary optic axis of the crystal then, as seen from the index ellipsoids of figure 2 , the birefringence can be reduced, and hence phase matching achieved, at longer wavelengths. Where the incident wave is a combination of ordinary and extraordinary polarizations, the phase matching is termed type I I .In this simple example it is apparent that there is a short-wavelength limit set by the birefringence and potentially, since the material i s transparent down to about 380 nm (as shown by the inset of figure 1) one might be able to provide shorter wavelengths if the birefringence could be increased, or if one could overcome the limitation set by the coherence length for a mismatched phase term. Practically, phase matching can be achieved if one rotates the crystal so that, using the index ellipsoids, the direction of propagation of the pump wave is at the correct angle with respect to the optic axis. Another way to achieve phase matching is to use the temperature dependence of the refractive index. Some limited flexibility exists for a chosen crystal as the indices have different temperature dependencies for the indices, hence heating, or cooling, can move the critical type I short wavelength limit. For example the temperature-dependencies of the refractive index are given by the Sellmeier equations for lithium niobate as

Electrlc fleld amolitude
I

I

.

I
I I

I

guide substrate I Figure 4. Transverse field profiles in symmetric planar waveguides for different mode values.
(9) where A is the wavelength in nanometres and T is the temperature in kelvins. So temperature tuning of the indices can allow phase matching. 3. Second-harmonic production in waveguides

-2.24 x ~ O - ~ A * 2.6 x 1 0 - ~ h ~

+

ni =

4.91 -k 1.173 x los 1.65 x 10-’T2 A* - (2.01 x i o 2 + 5.4 x 1 0 - 5 ~ 2 ) -2.78 x 10-8AZ

+

(8)

4.55 +0.97 x lo5 2.70 x 10-’T2 n3 = h2 - (2.01 x 102 + 5.4 x 10-5T2)2
1750

+

Identical problems exist in an optical waveguide as for a bulk crystal in that the material sets the fundamental values for the refractive indices. However, a guide structure offers some additional design flexibility because different modes propagate at different velocities through the guide since each mode has its own critically reflecting pathway in the

Periodicity poling for second-harmonic generation waveguide and is refracted with a different angle 0, at the boundary. Therefore. from Snell’s law the effective index of refraction depends on the mode of propagation. The index that corresponds to a wave whose wavevector is along the z axis (along the waveguide) is called the mode index n, (or effective index) and is found to be

(4

Air Non-dispersive glass (linear)

n,(w) = n(m)sinB,

(10)

where n is the refractive index of the medium and m is the mode number and 0 , is the mode angle. As shown in figure 3, the mode dispersion curves are a function of the guide width. Phase matching is satisfied at the intersection points of the pump and harmonic dispersion curves. The TE modes (field perpendicular to the direction of propagation and parallel to the plane of the waveguide) can be described (figure 4) by their electric field amplitude for a given frequency as

I

Barrier (low index) Crystal substrate

U
I

I

I

Fundamental (infra-red)

1

r

1

I
Sec. Harmonic (blue)

The behaviour of the field depends on the mode number and also fluctuates with its transversal position in the waveguide. For second-harmonic generation we are interested in the power density and hence how the fundamental power is converted to the second harmonic, and one important factor is the overlap integral between the fundamental and the second harmonic:

Cystal

This implies that the amount of energy converted from o to 2 0 depends on the modes selected since these will determine the value of A o v r , the overlap area. Assuming that the incident power is not heavily depleted and there is no reverse conversion, the efficiency q = PZ,/P, is easily deduced and

be achieved without problems o f cancellation of the overlap s integral. In this example !he phase-matching wavelength i for zeroth to second mode conversion [lS]. The modal dispersion curve as a function of guide width was sketched in figure 3 for wavelengths of 400 and 800 nm in quartz. From figure 3 it is apparent that the first requisite for SHG, namely index matching, occurs whenever there is a crossing on the dispersion curves. Three possibilities exist in this example for the zeroth mode of the fundamental to modes 2, 3 or 4 of the harmonic. Unfortunately this is only the first stage as the expression for SHG within a guide also requires evaluation of the overlap integral. The field profiles indicate that, whilst m = 0 to m = 0 would be ideal, odd to even overlap (such as m = 0 to m = 1) in a symmetric guide would give total cancellation, because the odd number modes contain equal quantities of positive and negative amplitudes in the field distribution. Solutions to this problem have been attempted and the simplest one conceptually is to form guides from a multilayer structure consisting of a mixture of nonlinear and linear material. In principle this would allow conventional mode propagation but inhibit the overlap integral in regions where intensity cancellations could be generated. Figures 5 and 6 sketch one such version of this suggestion [I31 in which the composite structure even allows mode matching at shorter wavelengths than would be feasible in bulk material. Despite the advantages of achieving SHG within a simple guide geometry, the fundamental problems of how to move to significantly shorter wavelengths and to select the optimal nonlinear coefficients still remain. In part, the first
1751

Figure 5. A composite waveguide structure formed with linear and nonlinear materials so that mode matching can

Pz, and P , are the harmonic and fundamental powers respectively, d and L are the nonlinear coefficient and length of the waveguide (this is the interaction length), n, and n b are the fundamental and second harmonic effective indices, EO is the vacuum permittivity, CO is the speed of light in the vacuum, A0 is the input wavelength and A,3 is the mismatch wavevector. The efficiency is optimized for sinc2(A,3L/2) being a maximum, which means that, in negative uniaxial crystals, the condition can be realized for the modal phase match in the same way as for a bulk crystal if n,(2o) = n&). Large values for the overlap integral imply a high SHG efficiency. For a step index planar waveguide the TE mode dependence of the effective index is given by

where d is the width of the guide, n , is the mode index; no the refractive index of air, nl the refractive index of the guide and nz the refractive index of the substrate.

M Houb and P D Townsend

Crystal (non-linear)

i

I

modes have continuously distributed propagation constants. Nevertheless the efficiency depends on the waveguide parameters, so an increase in the width usually implies a decrease in the efficiency. There are minimum values at specific depths, so the depth of the guide as well as the shape of the barrier are critical, as are the differences of the refractive indices between the guide and the substrate. It should be noted that the literature for proton-exchanged LiNbO3 waveguides is not totally consistent and, depending on the diffusion and subsequent annealing treatment, there are different opinions as to whether or not the guide region has nonlinear performance. Unfortunately, the Cerenkov light does not emerge from the sample in a well-defined axial beam and hence it is less convenient than a simple guided wave example of SHG.

4. Quasi-phase matching

0.0

0 . 8

1.6

As stated above, the second-harmonic wave is the resultant contributed by all the wavelets generated in the crystal. So for a laser of frequency w and wavenumber k l , propagating along the z direction, the field corresponding to the second harmonic is required to satisfy Maxwell’s equations with the boundary conditions:

Depth (PI Figure 6. Field profiles calculated from the model in figure 5 for the second harmonic (S,)mode 2, and the square of the fundamental (6) mode 0, together with the overlap evaluated over the nonlinear region [131. product (OVm) s for a barrierconfined waveguide 1.6 qm The calculation i thick (a) 100% LiNbOa and (b) a 50% layer of non-dispersive linear glass.
problem can be addressed by the use of Cerenkov phase matching, and both problems are amenable to the quasiphase matching (QPM) approach. In order to achieve SHG by using the Cerenkov radiation [14-18] there have been several experiments in LiNbO, waveguides, formed either by ion exchange or ion implantation.The goal of the operation is to obtain phase matching between a fundamental guided mode and a second-harmonic radiation mode (figure 7). That process can occur when the phase velocity of the guided wave is greater than that of the free wave in the substrate, which means nu c n b . LiNbOj is favoured because of its large birefringence (An = 0.11 at 0.8 j” and the low guide losses (0.2 dB cm-l). The efficiency for Cerenkov SHG from waveguides can be written as
q=
P z m =aLPu P ,

E

2x1

o ( eli(%~-2k~z)l

(16)
(17)

eli(br-k~z)l

J

ELd33E;dz

(15)

where a is a constant, L is the interaction length and d33 is the greatest nonlinear coefficient of the crystal. There exists an angle u,cosO = U%, which is the Cerenkov condition for phase matching. In the case of proton exchange, phase matching can be achieved at room temperature for an infinite number of radiation modes, so Cerenkov SHG relaxes the constraints on phase matching, since radiation 1752

with k2 # k , [IO]. Although, at z = 0, Eu,and Ek2 are in phase. because the waves travel with different velocities, they interfere destructively. Overall, the total harmonic intensity is proportional to IEu, Ek,I2 and is maximal for an odd number of coherence lengths le (minimal for an even number). For z < I, the second-harmonic power is built up from the fundamental and is at a maximum for z = I,. This corresponds to a phase shift of z.When z exceeds [1,-2L,] the power flows back to the fundamental. Figure 8 shows the Fresnel construction for the second harmonic. One solution to the cancellation problem is to construct an alternating succession of linear and nonlinear layers of length I,. Such a structure could be formed during crystal growth with two types of material. In a surface waveguide geometry it might alternatively be introduced by ion or electron bombardment to amorphize the crystal. The most commonly discussed alternative is termed periodic poling (figure 9(a)), which consists of a sequence of nonlinear segments of opposite optical domain direction. By changing the sense of the polarization vector a change of phase of ir can be produced for the secondharmonic polarization and therefore Eu, . If, for every odd multiple of I , such a reversal occurs, a phase shift of z is added to the existing one and destructive interferences can be avoided (see figure 9(b)). If N is the number of periods then the amplification expected is of the order of ( 2 N ) 2 1191. The second-harmonic polarization in the case of LiNbO3 is given by

+

P,(20, I) = 4,E:(w, t ) + d 3 1 E i ( ~ ,f)+d33Ei(w, t ) . (18)

Periodicity poling for second-harmonic generation Nonlinear polarisation field Fundamental

/

I
SHG.
FUNDAMENTAL

LiNb03 substrate

harmonic

I

Figure 7. Phase matching between a guide and a eerenkov wave for

,
I

I

I

Rgure 8. A Fresnel representation of SHG production from successive wavelets in which there is not perfect phase matching. 6 is the phase a t the source, which is the phase equivalent of z - s for SHG. Not only does this periodic domain inversion give fieedom from the necessity of index matching, but it can allow the use of coefficients with larger amplitudes. For example, the polarization of the laser light along the optic axis allows the use of d33 in the case of lithium niobate. lfSfor practical reasons, it is not feasible to build the periodicity on the scale required for optimal periodic domain reversal, then one could also use longer period inversions (figure 9(c)). In general the period for quasiphase matching (QPM) is given by overall the efficiency will be greater by a factor of 20 (see figure 9(b)) [21]. Periodic poling can be applied to waveguides and a theoretical treatment was given by Somekh and Yariv in 1972 using the Fourier expansion of d"(z) [22]. The efficiency q for the mode m = 0 is found to be

where A is the grating period and n is the QPM order. The efficiency relative to perfect phase matching is given by n ' (4/n2) (a first-order period is nine times more efficient than a third-order one) [ZO]. Tne factor 4/n2is off-set by the possibility of using higher nonlinear coefficients. For example in lithium niobate, d33 7d3 which means that

P ? and Pt are the powers for the harmonic and the fundamental for an m = 0 mode, de,, is the effective nonlinear coefficient, I the length of the crystal, W the width of the waveguide in the Y direction and f the thickness of the guiding layer. In principle, very high efficiencies may be achieved in a waveguide since W and f can be chosen to be very small (a guiding region which is only a few micrometres thick). However, in reality there are problems associated with coupling high pump powers into a guide, even though the subsequent confinement is
1753

M Houe and P D Townsend

(a)

1st order QPM

t - t - t - t - t - t -

-" C
2

h

I '
Petted Phase Matching

(i) introduction of a linear dielectric phase correcting plate, (ii) periodic insertion of crystals with an inverted crystalline orientation and (iii) changes in quality factor, Q. in a resonant structure to give a change of phase at each reflection. Thus the conversion efficiency would be increased by a factor of Q when the harmonic is in phase with the fundamental.

4 U

"

t

-

t

-

t

-

ideal for SHG, with high pump intensity being maintained, except for depletion into the harmonic. Note also that, for guides whose cross sectional dimensions are measured in micrometres the power densities within the guide are high (a power of just IO mW provides a power density of htw cm-Z). In comparisons of the conversion efficiency there are numerous experimental difficulties including uncertainties about the precise power coupled into the guide, reflections and guide losses. Therefore some authors quote efficiency as initially measured (%). others extrapolate to the performance expected if the input power were I W (%W), even though saturation or optical damage may occur at much lower power levels, and the projected numbers exceed 100%. Alternatively, one may allow for the effects of guide length and so quote a normalized efficiency (%W-' c&). Values quoted in this article are as given by the source articles and must be treated with caution.

5 . Polarization reversal
The idea of phase matching for SHG was first mentioned by Armstrong et nl in 1962 [9]. They considered the possibility of periodic polarization reversal for quasi-phase matching via a 180" shift of the relative phase to enhance the conversion efficiency. They proposed three possible techniques including:
1754

The ideas were developed by Miller in 1964, who gave a theoretical treatment of poling and experimented with it in BaTi03 [19]. The application of electric fields and heat treatment were described as empirically being potential poling techniques for a range of various materials [23,24] but rather less effort has concentrated on understanding the mechanisms [U]. The choice of materials for periodic structures has of course focused on those which perform well in normal bulk geometry. Materials suitable for SHG must be non-centrosymmetric, and consequently there is the possibility of reversing their spontaneous polarization with external help, either during growth or subsequently. Although quartz was the first crystal in which second-harmonic generation was observed, its nonlinear coefficients are not very large [26]. Larger values are found for barium titanate and so this material was of interest for early SHG experiments [27,28]: however, the production of barium titanate is both difficult and expensive. Similar production quality problems exist for many nonlinear crystals such as KNbO3 or KTiOPO4 (KTP). KTP is of current interest as it combines a Curie point at 936°C and a melting point at 1150°C. It has good electro-optic and nonlinear properties as well as a high optical damage threshold. The structure of KTP is composed of chains of Ti06 octahedra separated by PO4 tetrahedra and the crystal exhibits a z direction polarization [29]. KTP has been widely used for frequency-doubling YAG laser light [29] and periodic poling has also been reported [30-321. Despite the scale of the literature the crystals are still expensive and of variable quality. SHG performed on KTP crystals showed an efficiency of IO% for an incident pulsed signal of 1 p J energy and 20 os pulse length [33]. More recently, higher conversion efficiencies (> 30%) have been achieved by mode matching with waveguide structures both in W b O 3 1341 and in KTP [35]. Historically there was early interest in lithium tantalate. The domain structure has been studied by Antipov [36]and early reversal of the polarization was carried out by Ballman with the help of an electric field [37]. Periodic poling has been tried for example by Mizuuchi and Matsumoto, who obtained blue light by propagation in a waveguide [38401. Lithium tantalate is often resistant to photorefractive damage and presents large nonlinear susceptibilities (d33 = 26 pm Vi), and its useful transparency range goes down to 380 nm. Nevertheless, it is less commonly used than the isomorphous lithium niobate since the latter is more easily grown (owing to its lower melting point) and hence is more readily available. One of the most intensive crystal growth programmes has been devoted to lithium niobate and, because of this. high-grade material is readily available at a reasonable price. It also has the necessary large nonlinear

Periodicity poling for second-harmonic generation optical susceptibilities and electro-optic coefficients [41451. Nevertheless, it is susceptible to optical damage [46], in which high optical powers at visible wavelengths alter the refractive index and degrade the optical performance. This is particularly troublesome at shorter wavelengths and so may be a problem in SHG, even if there is no damage at the pump wavelength, as a high SHG conversion efficiency could generate photorefractive damage via the harmonic. The structure of LiNb01 is considered to be a distorted perovskite structure and, since we will focus on poling i n this material, it is useful to specify the structure. Lithium niobate is of the 3m class and space group R3c with a density of 4.64 g cm-’ at 25 “C, and the volume of a unit cell is 318.221 A’ (formula weight 147.843). The unit cell is described as either rhombohedral (two formula units per cell) or hexagonal (six formula units per cell) with cell parameters
aR

= 5.49

A

C~R = 55.52

A

aH = 5.15

A

CH

= 13.86 A.

The crystal is composed of distorted oxygen octahedra that are slightly rotated one above the others and have one face in common. The oxygen ions are arranged in a hexagonally close-packed fashion in planes perpendicular to the z axis. Those planes are separated by lithium and niobium ions at interstitial sites and the sequence Nb, vacancy, Li, Nb, vacancy, Li,. . .is situated along the optic axis ( c axis), whatever the temperature [5, 111. The Li+ and Nb5+ ions are much smaller than the 0’- ions (figure 10):

Figure 10. The structure of LiNbOa.

Rex-

= 1.40 A

R ~ i= i 0.68 A

RNb5t

= 0.70

A.

It can be seen that neither Li+ nor Nb5+ ions are situated at equal distances from the planes of oxygen [42], but instead, metal-oxygen plane distances are Li+-O- = 0.714 and 1.597 and Nb5+-02- = 0.897 and 1.413 The oxygen ions form triangles in planes perpendicular to the c axis and the greatest distance between oxygen ions in such planes is 3.362 A. Since the metallic ions can be displaced along the optic axis, their displacement is responsible for the ferroelectric or paraelectric state (the typical displacements of the LiC and Nb5+ ions from their closest plane o f oxygen are 0.45 and 0.25 A respectively). To go from the nonpolar paraelectric structure to the ferroelectric structure, the metal ions are displaced in the same sense and the direction of the displacement is responsible for the sign of the polarization. At the phase transition the Nb ions move to centrosymmetric sites whereas the Li ions mainly occupy random sites with some displacive elements [47]. Since LiNbO, has some weaknesses, particularly with respect to photorefractive damage, there are a variety of modified lithium niobate structures in which addition of Mg or Zn appears to reduce the photorefractive effect. There is also a search for alternative materials. For example, a-iodic acid (a-HIO]) belongs to a non-centrosymmetric space group with an orthorhombic unit cell. The crystal is negative biaxial and presents a strong nonlinear response: the nonlinear coefficient d14 is 1.5 times greater than d3l of LiNbO, and 20 times greater than dll of quartz. HI03

A

A.

does not suffer from optical damage at power densities of 2.5 x lo5 W cm-* and is easily grown. The transmission range is large: 4 W 1 3 0 0 nm. Other iodates such as KIO3, LiIO,, TI103 or RbIO, can also be used for bulk phasematched SHG. Unfortunately these materials are often easily decomposed [48,49]. No waveguide configuration or periodic polarization reversal has been carried out with them (to the best of our knowledge). Other inorganic crystals such as hydrogen-bonded dihydrogenphosphate have been studied by Aakeroy etal for their high resistance to optical damage and their transparency in the visible [50]. Urea, (NH&CO, is also very interesting for its nonlinear properties and its transparency in the W [51]. Phase matching has been produced at room temperature for A = 476 nm and second harmonics detected but here again, periodic poling has not yet been achieved. Many organic materials have been used for SHG, both in bulk and in waveguide structures but, even when the materials are non-centrosymmetric on the molecular scale, they may require treatments to produce poled structures. For example, they can be poled by application of an electric field above the glass transition and cooled under the field. Doping the polymer with active nonlinear molecules is done to form co-polymers such as poly(viny1edene cyanidelvinyl acetate) [52]. The crystals are highly polarizable because of their i ( electrons. The nonlinear properties of polymers are the subject of intensive research since their nonlinear response can be greater than that of inorganic crystals. Several molecules have been studied and poled, and polymers with poled side chains seem to be promising for electro-optic applications [53]. d,] coefficients of poled polymer glasses five times greater than those of KDP and SHG have been observed [54]. Quasi-phase matching in
1755

M Houe and P D Townsend

waveguides was tried by Nonvood in periodically poled polymers. Temperature, angle and wavelength tuning were carried out and an efficiency of 0.01% W-' was obtained for a fundamental laser light of 1.34 fim [ S I . Poled polymer multilayer waveguides have been successfully tried with a maximum efficiency of 0.4% W-' by modal dispersion [56]. One unfortunate problem is the limited life time of the poled structure. It has been noticed that organic molecules tend 10 lose their arrangement and thus their noncentrosymmetry. This limits the number of applications and so different techniques for poling polymers have been attempted. Anneser et al used a combination of light and electric field to pole glassy liquid crystalline polysiloxanes [57]. This technique can be undertaken at room temperature and should increase the life time of the poled structure. Very recently electron beam poling was tried by Yang et al and the experiment was carried out at 90°C for a primary energy of 10 keV 1581. The nonlinear coefficients obtained were insufficiently large for any application, but optimization of the technique would be of great advantage due to its simplicity of utilization. Organic salts are also prospective materials. 2-Methyl-4-nitroanidine (MNA) has shown a high second-harmonic response: its nonlinear coefficients d12 and dll appear to be 5.8 and 40 times greater than those of LiNbOs respectively 1591. Phase matching was possible by angle tuning, giving an efficiency 1.3 x IO6 times greater than that of quartz. 3-Methyl4-nitropyridine oxide (POM) has also been much studied and SHG has been detected [60,61]. Finally. one should note that some ceramics can be poled, but the process is quite difficult to achieve since each grain boundary has its own crystallographic axis, so one sixth of the domains will reverse in a tetragonal ferroelectric ceramic 151. More recently [62-651, it has been possible to pole silica glass by a combination of an electric field and a heated sample. The mechanism is unclear because the nonlinear coefficient is higher than expected for the formation of quartz. Optically driven poling is being actively studied within the cores of germano-silicate optical fibres. Yet again the models are tenuous, but probably involve germanium and impurity sites which catalyse the process 166,671. The preceding examples, and various reviews [5.68,69], indicate that there are a range of inorganic, organic and ceramic systems which are efficient for SHG and may thus be suitable for periodic poling. Table 1 gives an idea of the nonlinear coefficients of materials of current interest.
6. Poling m e c h a n i s m s for domain inversion

Table 1. Nonlinear coefficients of materials relative to KDP. Material d value/& (KDP) ADP RbHZPO4 KH~ASOA LiNb03 ' H103
Si02(quartz) BaTiOa KliOPO4
MNA KDP
&s=1 d m = 1.25 &=1.04 d.=l = 11.6 dia = 10.2

6;

Liloa

a1= 10.1
dil = 0.85
&I =36

POM

&6=15 d,, =464 &=14

can be destructive. Additionally, decomposition of surface layers can cause problems in waveguide structures. In many cases the materials are very sensitive to crystal field effects resulting from impurities or dopant ions and in simplistic approaches of poling by the application of large electric fields there may be long range ionic motion and electrical and structural breakdown, Polarization reversal was studied with barium titanate crystals by Miller and Savage as early as 1959 [70]. They deposited electrodes and utilized an electric field E applied parallel to the ferroelectric axis. Reversal was noted for an electric field of about lo3 V cm-'. Domain wall motion occurs during repoling and sideways wall motion over distances of about 1 mm was detected. In general, if the process is reversible, then the behaviour of the polarization as a function of the field follows a hysteresis loop and the coercive field necessary to reverse such a polarization can be as high as lo6 V cm-I. Barium titanate studies by M e n showed that there exists a current flow during reversal [71]. The maximum current observed follows the rule

= I,e-nlE

(21)

Poling of a ferroelectric can be induced by the application of a small electric field whilst cooling the crystal through the Curie point. This is a direct approach which is obviously useful for poling bulk crystals, either during or after growth. However, it may not be adequate for a controlled micrometre-scale periodic re-structuring of the domains, Therefore a variety of techniques are being considered. Associated problems of the materials mentioned so far can be that they dissociate at high temperature (even below the Curie point), or that they undergo phase transitions which
1756

where I , is the current for an infinite field and (Y is a constant dependent on temperature and sample thickness. A possible mechanism was given in which nucleation of antiparallel domains occurs near the surface; this is followed by both forward and sidewise motion of the walls. Questions arise as to how domains nucleate and how they expand. Factors discussed include the existence of space charge layers near the surface which induce electric fields. The strength of space charge fields is probably not high enough to generate the inversion process directly, but it might contribute when an extra field is applied externally. In this example the growth of domains is mainly governed by domain wall motion. To explain how domains can expand, energetic and kinematic studies of reversal have to be considered. The velocity of the side wall movement was found by Miller [72] to be independent of the domain size. Electrical pulses were also observed when electric fields were applied. These pulses were caused by the Barkhausen effect, due to jerky motion of the domain walls [73,74]. Barkhausen-type pulses were attributed to the nucleation mechanism [72,74,75] or to sideways motion [76]. It

Periodicity poling for second-harmonic generation has been suggested that sideways motion of the domain walls occurs through a succession of small triangular steps along the wall [77-791. It has also been submitted that the energetic state of front walls is much higher than that of side walls, resulting in growth that is mainly dependent on the applied field, and a small contribution from the depolarizing field. Crystal imperfections undoubtedly play a significant role in any such restructuring and the influences may come from atomic-scale point defects, such as vacancies or impurities, or from extended defects such as surfaces or dislocations. A significant problem in separating the many factors involved is that the quality of the crystals is highly variable in most cases, and so the dominant factors may be sample dependent. In particular, screw dislocations have been identified as inhibitors of wall motion. Nakamura, in 1954, postulated the existence of screw dislocations contributing to the wall movement process in BaTi03 [SO]. When the Burgers vector is perpendicular to the domain boundary the border is a helicoidal surface. The step formed by the intersection of this surface and the dislocation line can move along the boundary and after several revolutions a movement of the boundary can be produced. Unfortunately the fact that there are several dislocations in the crystal implies that the step does not move freely. Miller and Weinreich concluded that, by considering the screw dislocation model, the approximate form of the wall velocity would be
F(E)c ( e-XW2/(EzP:)

(22)

Their theoretical treatment contradicts the experimental wall velocity which is of the form e-Mnsmnt/E, Perhaps. with a very high concentration of dislocations, the sources are not active, even at high applied fields, and the possibility remains that the thermal energy of the material might excite some sources to become active centres for nucleation [78]. It has been assumed that several steps can nucleate simultaneously in order to explain that the wall velocity is not proportional to the wall length. Hayashi, in 1971, proposed that this wall velocity is due to a change in step nucleation [SI] (further information is available in the article written by Stadler [U]). Barium titanate presents a first-order transition between the ferroelectric and paraelectric states; thus such explanations may need re-appraisal for most other ferroelectric materials, which show second-order transitions.
7. Poling techniques

7.1. Electric fields and electron irradiation

Bulk and periodic poling in LiNbO3 has been attempted via a wide number of treatments involving chemistry, heat and electric fields. In many cases these are empirical methods in which the end result is useful and, realistically, there is little understanding of the details of the process. Reference to the structure of lithium niobate in figure 10 shows that inversion of the poling direction will result if the lithium ion can be induced to move across the plane of the triangle of oxygen ions. At the Curie temperature the average Li position

moves into the plane, as the phonon modes soften the lattice structure. At lower temperatures large electric fields, of either external or chemical origin, could force the lithium through the triangle. However, the requisite fields must be pulsed because permanent application of sufficiently high fields will cause permanent breakdown of the lattice and it is obvious that, since heating allows relaxation of the lattice, a smaller field is needed at higher temperatures than at room temperature. The requisite fields are usually pulsed to avoid permanent breakdown. However, if highquality crystals and electrodes are used, field effects at the edges of the samples can be avoided. Poling was recently reported in 0.5 mm thick LiNbO3 using highvoltage fields for a duration of up to 1.9 s [83]. Camlibel used a pulsed DC field to induce polarization reversal [25] and avoid lattice damage and, in 1972, Ballman and Brown experimented with a pulsed field technique to invert single-domain lithium tantalate crystals. The samples were heated to different temperatures and a pulsed DC field was applied across the c-cut specimens. Hence the required poling conditions were established and it was noticed that the field required decreased with increasing temperature (figure 11). For a field of IO4 V cm-I applied at 250"C, an 80% domain reversal was observed after 20 min but it took about 4 h with a field of 5 x IO3 V cm-' under the same conditions and reversal did not occur at all for a field of IO3 V cm-I [39]. The electric field method has been applied to periodic poling, for example, LiTaO3 being periodically reversed by electrodes applied on its positive c face when a sample was heated to 600°C (i.e. close to the Curie temperature) and 1.4 V applied for IO min. Bulk LiNbO3 (0.14.2 mm thick) has been successhlly poled at room temperature and with periodic liquid electrodes under a field higher than the coercive field [84]. A recent application to SHG has shown an efficiency of 55% for an input power of 55 mW (Webjorn, private communication). Third-order QPM was tried in planar and channel waveguide configurations to obtain efficiencies of 0.2 and 8% W-' respectively [85]. In another experiment, aluminium electrodes deposited on both faces of LiNb03, with a periodic pattern drawn on the +c face, allowed poling at room temperature with negative pulsed voltages applied on the -c face, while the t c face was earthed. This induced an electric field of about 24 kV mm-'. Domain inversion occurred under the electrodes on the +c face and the domain walls observed were parallel to the optic axis. Similar SHG experiments were carried out in proton-exchanged waveguides, in which a normalized efficiency of 600% W-l was realized [86]. The width of the electrodes can control the SHG performance and a maximum efficiency was observed for the wide electrodes [87]. An alternative way to initiate an open pathway for Li motion was considered by Townsend in 1977. He proposed that ionization within the crystal (by fast electrons or xrays) could locally excite the crystal lattice and generate a transient molecular oxygen ion. The key feature of this suggestion is that the molecular ion is physically much smaller than the original pair of independent oxygen ions (ozone equivalent states may also exist). This relaxation
1757

M Houe and

P D Townsend

electron beam

+ face

Figure 12. An example of the shape of the inversion region in LiNbO:, which resulted from electron beam bombardment of the + surface. The inversion region i s much deeper than lhe electron penetration depth. the thickness of the sample, which can be 100 times greater than the electron penetration depth. Models which might explain this could simply be that the local electric field is well above that needed for inversion but, because of beam-induced conductivity there is a conducting plasmatype region which moves the potential boundary deeper into the crystal. Beam heating will exist but except for an initial local surface heating the power levels are unlikely to raise the bulk temperature a significant amount. Nevertheless pyroelectric forces may assist the relaxation and domain wall motion. In a typical scanned electron beam experiment the electron beam is scanned on the uncoated -c face of the crystal with adotted line mode. Scan rates of 30-50 @m s-' are used for dot spacings of 2-4 k m . The f c face is coated and earthed. 2G!5 keV primary electrons are generally used (but poling has been noted even for 2 keV beams 1981). The grating period achieved at the higher energies is 4-7 p m [90-921 (figure 12). An advantage of the technique is the use of the electron beam for both the grating drawing and the field production. This method leads to 180" domain walls 1911 (figure 13(a)). Waveguide applications produce high second-harmonic conversion efficiencies of up to 50% W-' [92J. By increasing the scan rate to 250 p m s-' an efficiency of 70% W-' has been realized [93]. Other materials, such as lithium tantalate and KTP, have been successfully tried 132,941. It seems that it is easier to reverse domains in KTP and an SHC efficiency of 7 x IO-' has been produced for a fifth-order QPM [32]. Optimization of the beam and crystal parameters will be possible once the mechanisms are understood.

12.71

4.3

i27

237 T (deg C)

363.6

721

Figure 11. Data for effective polarization reversal in LiTa4 as a function of temperature and DC field [37].Note that region A is stable against inversion whereas excursions
into region B allow re-poling. creates a vacancy in the oxygen triangle and a large pathway for the lithium ion to cross to the other side, whilst under the influence of a small electric field. Since the saddle point is at sub-AngsWm unit distances from the oxygen plane it only requires that the oxygen molecular state be metastable on the picosecond time scale. In experiments to test this idea crystals of lithium niobate and tantalate were re-poled at temperatures as low as 600 and 400°C [88] (well below the Curie temperatures). In an attempt to achieve the same result in a waveguide structure, a periodic region was inverted by I O keV energy electrons scanned into the crystal through a mask pattern [89] in the presence of a small electric field (10 V cm-'), with the lithium niobate at 580°C. Inversion on either the +c or the -c face was recorded by suitable choice of the direction of the electric field. Subsequent experiments [90-941 have used electron beams from a scanning electron microscope to write a periodic pattern directly, and so avoid the necessity of a metallic mask. However, the method does not allow application of an external electric field but is only applicable to the negative c face of the crystal. In this case the electron beam both generates a space charge beneath the surface and causes emission of secondary electrons. The surface potential must therefore approach a condition defined by unitary secondary electron emission and equilibrium charge leakage. This can be several kilo-electron-volts and, for IO keV electrons, the field may fall over sub-micrometre distances, hence very high fields can develop. This model gives a possible explanation of how an apparently field free situation can induce poling on the -c face (but not on the +c face). The details of the development of the field are complex and Cazaux has studied several such situations [95-971. The speed of poling depends on the current density, as one might expect, but a more surprising observation is that domain inversion steadily extends across
1758

7.2. Heat treatments
Heating towards the Curie temperature can mimic the simple classic conditions for poling but if the region is to be localized, as in periodic poling, then bulk heating may be undesirable. A recent variation of the Ballman approach defined by figure 11 was to apply a steady DC electric field to crystals of lithium niobate and pulse the local temperature by using a laser beam of 5 ns duration and some 80-100 mJ cm-* 1981. The concept is to move from the stable regime defined below the boundary line (of figure 11) to a point above it. Whilst a high-voltage

Periodicity poling for second-harmonic generation sample temperature was 400°C and the steady field was just 750 V cm-' (figure 13).

1.3. Chemically linked poling effects
Heat treatments of lithium niobate and the presence of contacts may result in chemical changes which alter the ease of poling, or actually induce it. It should be noted that, at high temperature, lithium niohate partially dissociates and can enter into a wide range of related Li-Nb oxide phases [99]. This has been exploited in several poling experiments, for example, when uncoated c-cut samples were heated in a quartz tube furnace for several hours and cooled down to room temperature at a rate of 50°C min-' in an atmosphere of air or Ar gas containing water vapour, polishing and etching revealed that inversion had occurred. It has been found that the thickness of the inverted layer increases with the heat treatment time and the cooling rate. The L i z 0 out-diffusion and the pyroelectric field induced during cooling may contribute to the inversion process. The inversion depth was about 100 p m for a heating time of 4 h hut could reach 250 p m after I O h. It has been shown that the nucleation started on the +c face of the specimen [1001.
7.4. Li20 out-diffusion

Figure 13. A comparison of domain boundaly pattems in the case o f LiNbO, as a result of different poling treatments. (a) 180" domain reversal from electron beam irradiation: (b) triangular domain boundaries at 30" after titanium diffusion and (c) semicircular domain boundaries in proton exchange surfaces.

Poling can be induced on the +c face of LiNbOz when striped bands of silicon oxide are deposited at the surface of the sample that is heated to temperatures close to T, for a time of about 90 min. Liz0 out-diffusion occurs for uncovered parts and induces a change in the composition at the surface due to a Li deficiency in the material. This probably lowers the Curie temperature, which results in inversion [loll. Grating periods of 2-3.5 p m have been made this way. The shape of the inverted zones is triangular with 30" domain walls and the typical depth observed is 1.5-2 p m (figure 14(b)). In waveguides the conversion efficiency is about 0.7% W-', which is much lower than in some other procedures [ l l ] , although a conversion efficiency of 6% W-' has been produced [102].
1.5. Surface chemistry

Figure 14. Polarization reversal as the result of a thermal pulse from a laser beam, which heated t h e LiNbO. from the equivalent of region A to region B (figure 11). The circular region had been coated with copper to establish a DC voltage level. This has mostly inverted in polarity in the region where the laser was incident except for a small dot (A) and an irregular patch (8).(C) shows an indirectly illuminated area without copper, where inversion has also occurred [98].

pulse moves between conditions in the ordinate direction the heat pulse move alsng the abscissa of figure 11. In this first example domain inversion was demonstrated when the

Changes in the crystal field induced by chemical reactions or substitution of impurities can cause domain reversal. Several such examples have been documented, some of which are associated with techniques used in production of lithium niobate waveguides. Reference will now be made to domain inversion resulting from proton exchange, silica cladding and reactions with titanium films. Proton exchange followed by heat treatment produces a reversed pattern on the +c face of LiNbOz and the --c face of LiTaO, only. The method has been to deposit a Ta mask, pattern with CF, dry etching, then proton exchange in a bath of pyrophosphoric acid at temperatures of 220260°C for 2&50 min. The Ta mask is removed by NaOH or H F etching. This process is followed by a heat treatment at 5 5 M 0 0 " C for about I O min and semicircular inverted domains were observed with an inversion depth in the range 1.7-3 Wm (figure 14(c)). This type of proton exchange has
1759

M Houe and P D Townsend

mainly been carried out in LiTaO3 substrates [38,1021 and this method has the advantage o f being performed at low temperatures and without a mask during the heat treatment. It has recently been shown that the inversion depth increases with the heat treatment temperature and that increasing the proton exhange time lowers the threshold temperature for domain reversal 1381. Maximum conversion efficiencies obtained in waveguides are about 157% W-' for a firstorder QPM [IO31 and 27% W-' for a third-order QPM 1401. When an Si02 film is deposited on the +c face of the substrate and the surface is patterned either by chemical etching or by lift-off Lithography, followed by a heat treatment near Tcfor several hours [104], inversion occurs under the coated areas. The results obtained were an inversion depth of 1.3 wm for a grating period of 7 pm, which is suitable for waveguide applications. The shape of the inverted regions is triangular (domain walls at 30"). There is an upper limit in temperature, 11OO"C, above which the coating cracks and the inversion occurs across the whole surface. A waveguide application has shown a second-harmonic normalized efficiency of 16% W-' . Thermal oxidation of Ti films on LiNb03 crystals at about 500°C minimizes the refractive index changes whilst forming inverted regions. The reversed domains appear on the -c face, outside the titanium oxide pattern. The depth of inversion is approximately 2 p m for a grating period of IO p m , This method has the advantage of operating well below Tc. When proton-exchanged waveguides were formed along the X stripes the normalized SHG efficiency produced was 1.6% W cm-' [ 1031. T i in-diffusion is routinely used to form waveguides in LiNb03. The method usually involves the deposition of a Ti film at the surface of the specimen and lift-off lithography is used to pattern the sample. Ti in-diffusion occurs when the crystal is heated to 900-1100"C for 510 h and inversion occurred at 1020°C [105-107]. The process leads to a domain reversal only on the +c face for the non-diffused parts of the surface. The higher surface concentration of Ti on the +c face could explain this phenomenon because new Tifniobate structures are known to exist. The typical inversion depth is of the order of 2-4 p m and it has been shown that the inversion depth is proportional to the Ti thickness [ 1081 and the diffusion time [ZI]. The values of the electro-optic coefficients were found to be reduced by domain inversion [108]. The shape of the inverted regions is triangular with an angle of 30" with the surface boundary (figure 14(b)). Insofar as the applications in waveguides are concerned, the first experiments for SHG lead to normalized efficiencies of about 37% W-' cm? [log, 1101. More recently, the same experiment performed with KTP showed an efficiency of 50% W-' cm-' [31]. Simultaneous blue (433 nm) and green (532 nm) lights have even been obtained in LiNbO3 channel waveguides with conversion efficiencies of 22 and 28% W-' respectively [ I 1I]. This method, though efficient, increases the risk of photorefractive damage in waveguides. Furthermore, the shape of the inverted regions is not ideal for QPM applications. In the case of Ti in-diffusion, a possible explanation has been given by Thaniyavarn [log] in which the Ti surface
1760

concentration is increased due to its lower diffusivity at the +c face of the crystal. This change of concentration increases the susceptibility of the material and lowers the Curie temperature. Since in-diffusion is carried out at temperatures close to T,,this might be sufficient to reverse the polarization and explain why reversal only occurs at the +c face. Liz0 out-diffusion implies a change of the composition near the surface, probably for the same reasons as Ti in-diffusion. Since the Curie temperature depends on the Li ratio in LiNbO3, the Curie point is also lowered. This mechanism might apply more generally for the heat treatment methods.

7 . 6 . Miscellaneous methods
Laser-heated pedestal growth with 10.6 @m radiation focused on the tip of a 250 p m diameter single-crystal rod of LiNbO3 can be made to give periodic inversion during growth in air by periodically modulating the heating power. For a grating period of 2.7 prn, and first-order QPM, there was an efficiency of 0.52% W-'mm-' \*.hen using the d33 coefficient [112]. On coating a planar waveguide with a periodic layer of a nonlinear organic material, such as a film of dipoly(methyl)methacrylate, a gating could be produced with a YAG laser. The period was large at 10 p m but the conversion efficiency obtained was low at 1.1 x for an input power of 25 kW [ 1131. Corona effects can result in periodic poling of both lithium niobate and polymers. The mechanism may resemble the earlier process of electric field or electron beam bombardment. Table 2 gives some reults
8. Methods for monitoring domain

reversal

Whilst the current objective is for SHG it is obviously useful to have techniques by which to sense whether domain reversal has occurred, preferably by a non-destructive route. For bulk crystals a pyroelectric charge develops when a change of temperature is applied and this determines the sign of the polarization. All the ferroelectric crystals are pyroelectrics and hence possess such a characteristic. This can be written as P = piAT (23) where P is the electric polarization, pi is the pyroelectric coefficient and AT is the applied change of temperature. The temperature gradient results in an excess of free charges at the surface of the material, resulting in the formation of a voltage. The same kind of measurement can be performed for these piezoelectric materials as the polarization is changed when a stress is applied via a metal contact. Typically a delicate impact will generate an observable charge and voltage pulse. The most utilized technique is chemical etching of the LiNbO3 specimen (but this is a destructive measurement). It has been shown that a solution of 1 2 HFHN03 reveals the domain walls. Effectively, it etches the -c face quicker than the +c face. Most of the time it is used at its boiling point (1 10°C) for a few tens of minutes or, in the case of proton exchange re-poling, at room temperature for 1 h,

Periodicity poling for second-harmonic generation

Table 2. Results obtained with various materials and techniques.
Methods Electric field Material LTa03(+c) LiTa03(+c) LiNbOa(+c) LiNbOs(-c) LiNbOs(-c) Electron-beam KTP (-c) LiNb03(-c) LiNbO$(-c) LiNb03(-c) LizO out-diffusion Proton exchange LiNbOs(+c) LiNb03(+c) Lira03(-c) LTa03(-c) LiTa03(-c) LiTa03(-c) LiTaO,(-c) LiTa03(-c) Fifth First First First First Third Third First Third Third First First First Third Order Efficiency q
8% W-’ cm-2
0.2% W-’ cm-2

PO
41 mW
20.8 mW

P2”
1.3 W IL 7.5 nW 20.7 mW
20 &W

Conditions
li:Al,Oa .. I = 1 mm chan neI TkAl203 / = 1 mm planar li:A1203 i=5mm RA1203 /=3.3 mm YAG / = 4 mm

Reference

600% W-’ cm-2 0.07% W-’ cm-I 55%
7 10-5

196 mW
300 mW 55 mW 139 mW
20 mW

1.3 &W 3.5 &W WG:-I=1.5mm YAG

40% W-’ c m 2
50% W-‘

70% W-’ 0.7% W-’ 6% W-’ 18% w-’ 10.3% 27% W-’

Ti:Al,O,
WG: I= 3.3 mm

3 mW
5 mW 26.6 mW 145 mW 210 mW 42 mW
121 mW

60 nW 1.5 NW 0.13 mW 15 mW 12 mW 0.32 mW 23 mW 1.1 mW

Stywl-9 dye WG: / = 1 4 m m li:A1203 WG Ti:A1203 WG:1=9mm li:A1203 WG: I= 9 mm
lIA1203

18% w-’
157% 3%

37 mW

WG:i=9mm Ti:A1203 WG TkA1203 WG:/=lOmm GaAlAs diode WG:/=lOmm YAG WG: / = 7 m m YAG WG:1=6mm YAG WG: I = 6 mm YAG WG Styryl-9 dye WG:l=7mm YAG WG:l=4mm

Ti in-diffusion

LiNbOs(+c) LiNb03(+c) LiNb03(+c) LiNbO$(+c) LiNbOs(+c)

First First second Third Third

0 . 1 % W-’ cm-2
28% w-

800 mW

360 &W
0.25 mW

30 mW

22%

w-‘
1 mW 14.7 mW
0.4 mW

5% W-’ c m 2

(green) 0.45 mW (blue) 0.5 nW
940 nW
26 nW

37% W-’ cm-2 16% W-’

SiOn cladding Laser-heated pedestal

LiNb03(+c) LiNbOs First

8

10-5

83 mW

6.7 W

Stytyl-9 dve Bulk

which i s more convenient and gives more reproducible results. However, with other techniques the etching time can be as long as 30 h. The etchant is material-specific and a solution of two parts of KOH for one part of KNO3 used at 220°C for 2 s has been successfully used on KTF’ (also on LiNb&), but with different etching rates. The results are often more reproducible and smoother than with

E W . A solution of 1:l CsOH:KNO3 also works, albeit less efficiently. The visualization of the domain boundaries can be made with a microscope, for example. X-rays can also be used because the twins at the crystal surface that occur during polarization reversal can be depicted. I t has been shown that the intensity of the (0006) reflection differs from the (0006) one for LiNb03.
1761

M Houe and P D Townsend

Since poling involves a change in these two directions, a difference of contrast can be observed. The positive domain intensity is greater than the negative domain intensity by a factor of 1.271. X-ray scattering also shows the domain boundaries of doped K T F ' with either K or Rb, which induces domain reversal. Toning has been reported to be a possible method for detecting inverted regions by using the pyroelectric effect at low temperatures: drops of liquid electrostatic toner diluted with either positive or negative particles are deposited onto the cooled sample (-77°C for KTP). Positive (negative) particles are attracted by -c(+c) orientated domains and domain boundaries can be observed. U V photoemission is another possibility since the photoelectric yield changes from one domain to another. Similarly. the use of a scanning electron microscope can show the domains. This is possible with modern scanning microscopes which can operate at voltages as low as 2 ' kV and do not need metallization on insulators. The 2 kV value is close to the equilibrium point for unit secondary electron emission, but the local field from and - faces represents a sufficiently large potential difference that there is good contrast between the domain directions. Figure 13 shows regions of a surface which have been re-poled by laser pulses and which have been visualized by this technique 1981. Note, however, that it is essential to use low beam currents or the electron beam will itself induce domain reversal.

References
[I] Franken P A , Hill A E, Peters C W and Weinreich G 1961 Phys. Rev. Lett. 7 I 1 8 [2] Kleinman D A 1962 Phys. Rev 128 1761 [3] Conwell E M 1973 IEEE J. Qurnt. Electron. 9 867 [4] Baldwin GC 1974 An lntroduction to Nonlinear Oprics (New York: Plenum) [5] Lines M E and Glass A M 1979 Principles and ADDlications of Ferroelectrics and Related Materials .. (Oxford Clkndon) [6] Shen Y R 1984 The PrincipIes of Nonlinear Optics (New York Wiley) 171 d. . . SHG Bratz A, Felderhof B U and Marowskv G 1990 A.. Phys. B 50 393 181 Yariv A 1991 OpticaIElecrronics 4th edn (Philadelphia: Saunders CoUeze) .., [9] Armstrong J A, Bloembergen N, Ducuing J and Pershan P S 1962 Phvs. Rev 127 1918 [IO] Lim E J. Heriz H M, Bortz M Land Fejer M M 1991 59 2207 11 I] Prokhorov A M and Kuz'minov Y S 1990 Physics and Chemistry o f Crysralline Lithium Niobate (Bristol: Adam Hilger) [I21 Kirby C J G 1989 Properties of Lirhium Niobate INSPEC EMlS series vol 5 , p 156 [I31 Townsend P D, Zhang L and Chandler P J 1994 J. Mod. Opt. 41 1349 1141 Li M 1. De Micheli M, He Q and Ostrowsky D B 1990 IEEE J. Quant. Electron. 26 1384 [I51 Keys R W, Lon1 A and De La Rue R M 1990 J. Mod. Phys. 37 545 [I61 Chikuma L and Umegaki S 1990 J. Opr. Soc. Am. B 7 768 [I71 Hashizume N. Kondo T, Onda T. Ogasawana N, Umegaki S and It0 R 1992 IEEE J. Quant. Electron. 128 1798 [IS] Hamelin N, Lifante G, Chandler P J , Townsend P D, Pityana S and McCaffrey A J 1994 J. Mod. Opt. 41 1339 [19] Miller R C 1964 Phys. Rev. A 134 1313 [ZO] Webjbrn J, Laurell F and Arvidsson G 1989 IEEE Photon. Tech. Lett. 1316 [21] Webjam J, Laurell F and Arvidsson G 1989 J. Light. Tech, 7 1597 [22] Somekh S and Yariv A 1972 Opt. Co,nmun. 6 301 [23] Savage A 1966 J . Appl. Phys. 137 3071 [24] Camlibel I 1969 J. Appl. Phys. 40 1690 1251 Hobden M V 1967 J. Appl. Phys. 38 4365 [%I Babsail L. Lifante G and Townsend P D 1991 Appl. Phys. k r t . 59 384 [27] Chynowelh A G 1956 J. Appl. Phys. 27 78 [28] Nakao 0, Tomomauu K, Ajimura S,Kurosaka A and Tominaga H 1992 Appl. Phys. Lert. 61 1730 [29] Bierlein J D and Vanherzeele H 1989 J, Opr. Soc. A m B 6 622 1301 Vanheneele H and Bierlein J D 1992 Opt. Lett. 17 982 (311 Van Der Poel C J, Bierlein J D and Brown J B and Colak S 1990 AppL Phys. Letr. 57 2074 [32] Gupta M C, Risk W P. Nutt A C G and Lau S D 1993 Appl. Phys. Lev. 63 I167 1331 Zhang L, Chandler P J, Townsend P D, Alwahari Z T and McCaffery A J 1992 Electron. Lett. 28 1478 1341 muck D. Binder B. Looser H. Buchal Ch and Giinter P 1992 Opr. Commun. 90 304 [351 Zhang L, Chandler P J , Townsend P D. Alwahabi 2 T, Pityana S L and McCaffery A J 1993 J. Appl. Phys. 73 2695 [361 Antipov VV, Blistanov A A, Sorokin N G and Chizhikov S Phys.-Crysrallogr. 30 428 1985 SOP. 1371 Ballman A A and Brown H 1972 Ferroelecrrics 4 189 i38j Mizuuchi K. Yamamoto K and Taniuchi T 1991 Appl. Phvs. Lett. 58 2732 1391 Mizuhchi K and Yamamoto K 1992 Appl. Phys. Lett. 60 1283

+

9. Conclusion
In this paper, periodic poling has been discussed as a valuable technique to produce high output conversion efficiency in SHG configurations via QPM. The advantazes have been clear for 30 years and, although the discussion has primarily been for lithium niobate, it can be applied to a wide range of materials. Nevertheless, whilst poling can be empirically achieved, the underlying mechanisms are far from being well established. Therefore in order to improve the processes for the niobate, and to examine which routes are more generally applicable, it is essential to understand the basic mechanisms for poling. Even now one cannot uniquely explain the shape of the inverted domain, and examples of both triangular or semicircular structures are observed. Also, one is unsure why the electron beam methods are, at the moment, the only way to produce 180" domain walls. Nevertheless, quasi-phase matching remains a very promising technique for integrated optics applications and the commercial incentives to give controllable periodic poling ensure that this topic will be a key problem in photonics for several more years.

Acknowledgments
We very much appreciate critical reading of the manuscript by Drs Gines Lifante, Jose Olivares and Jonas Webjorn. We also wish to thank the EPSRC for financial support.
1762

Periodicity poling for second-harmonic generation

[a] Mizuuchi K and Yamamoto K 1992 J. Appl, Phys. 72 5061 I411 Abrahams S C, Reddy J M and Bemstein J L 1966 J. Phys. Chem. Solids 27 997 [42] Abrahams S C, Hamilton W C and Reddy J M 1966 J. Phys. Chem. Solids 27 1013 [43] Abrahams S C, Levinstein H J and Reddy J M 1966 J. Phys. Chem. Solids 27 1019 [44] Boyd C D, Miller R C, Nassau K, Bond W L and Savage A 1964 Appl. Phys. Lett. 52 34 [45] Niizeki N, Yamada T and Toyota H 1967 Japan. J. Appl, Phys. 6 3 18 [46] Chen C Y, Sweeney K L and Halliburton L E 1984 Phys. Srntus Solidi a 81 253 [47] Boysen H and Altorfer F 1994 Acta Cysrallogr. B 50 405 [a] Aakeroy C B 1990 DPhil Thesis Sussex University [49] Kurtz S K, Perry T T and Bergman J G 1968 Appl. Phys. Lett. 12 186 [SO] Aakeroy C B, Hitchcock P B, Moyle B D and Seddon K R 1989 J. Chem. Soc. Chem. Commun 1856 [51] Cassidy C, Halbout J M, Donaldson W and Tang C L 1979 Opt. Commun. 29 243 1521 Kurihara S , Takahashi Y, Miyaji H and Seo I 1989 Japan. J. Appl. Phys. 28 L686 [53] Van Der Vorst C P J M. Van Rheede M and Van Weerdenburg 1993 Int. J. Polym. Mater. 22 113 [54] Singer K D, Sohn J E and Lalama S J 1986 Appl. Phys. Lett. 49 248 1551 Norwood R A and Khanarian G 1990 Electron. Lett. 26 2105 [56] Rikken G L J A, Seppen C J E, Staring E G J and Venhuizen A H J 1993 Appl. Phys. Lett. 62 2483 [57] Anneser H, Feiner F, Petri A, Bauchle C, Leigeber H, Weilzel H P, Kreuzer F H. Haak 0 and Boldt P 1993 Adv. Mater. 5 556 [58] Yang C M. Bauer-Gogonea S and Sessler G M 1994 Appl. Phys. Len. 64 22 [59] Levine B F,Bethea C G , Thurmond C D, Lynch R T and Bemslein J L 1979 J. Appl. Phys. 50 2523 [60] Sigelle M and Hierle R 1981 J. Appl. Phys. 52 4199 [61] Hierle R,Badan J and Zyss J 1984 J. Cryst. Growth 69 545 [62] Myers R A. Mukherjee N and Brueck S R J 1991 Opt Lett. 16 1734 I631 Okada A, Ishii K, Mito K and Sasaki K 1992 Appl. Phys. Lett. 60 2853 [64] Mukherjee N, Myers R A and Brueck S R J 1994 J. Opt. Soc. Am. B 11 665 [65] Dominic V and Feinberg J 1994 J. Opt Soc. Am. B 1 1 2016 [66] Kashyap R I989 J. Opt. Soc. Am. B 6 313 [67] Kazansky P G. Pruneri and Russell P S l J 1995 Opt, Lett. 20 843 [68] Kolinsky P V and Jones R J 1989 G E C J . Res. 7 46 [69] Fejer M M, Magel G A and Lim E J 1989 Proc. SPIE 1148 213 [70] Miller R C and Savage A 1959 Phys. Rev. Lett. 2 294 [71] hlerz W J 1956 J. Appl, Phys. 27 938 [72] Miller R C and Savage A 1958 Phys, Rev. 112 755 [73] Newton R R, Aheam A J and McKay K G 1949 Phys. Rev. 75 103 [74] Chynoweth A G 1958 Phys. Rev. 110 1316

[75] [76] [77] [78] [79] [SO] [SI] [SZ] [83] [84]

Chynoweth A G 1959 J. Appl. Phys. 30 280 Miller R C 1958 Phys. Rev. 111 73 Abe R 1959 J. Phys. Soc. Japan 14 633 Miller R C and Weinreich G 1960 Phys. Rev 117 1460 Nakamura T 1954 J. Phys.'Soc.Japan 9 425 Miller C and Savage A 1960 J. Appl. Phys. 31 662 Hayashi M 1972 J. Phys. Soc. Japan 33 616 Stadler H L 1992 Ferroelectrics 137 373 Myers E L 1995 Opr. Lerr. at press Webj6m J. Pruneri V. Russell P ST J. Barr J R M and Hanna D C 1994 Ekcrron Lett. 30 894 I851 Matsumoto S, Lim E I, Hertz H M and Fejer M M 1991 Electron. Len 27 2040 [86] Yamada M. Nada N, Sailoh M and Watanabe K 1993 AppL Phys. Lett. 62 435 [87] Brauer M L, Dajani I and Kester J J 1994 Electron. Lett. 30 297 [88] Haycock P W and Townsend P D 1986 Appl. Phys. Lett. 48 698 [89] Keys R W, Loni A, De La Rue R M, Ironside C N, Marsh J H, Luff B J and Townsend P D 1990 Electron Letr. 26 189 [90] Yamada M and Kishima K 1991 Electron Lett. 27 828 [91] It0 H, Takyu C and lnara H 1991 Electron. Lett. 27 1221 [92] Fujimura M, Suhara T and Nishihara H 1992 Electron. Lett. 25 721 [93] Fujimura M, Kintaka K, Suhara T and Nishihara H 1992 Electron. Len. 28 1868 [94] Hsu W Y and Gupta M C 1992 Appl. Phys. Len. 60 1 [95] Cazaux J 1986 J. Appl. Phys. 59 1418 [96] Cazaux J and Le Gressus C 1991 Scanning Microsc. 5 17 [97] Cazaux J 1992 J. Electron. Spectrosc. Relar. Phenom 59 1418 [98] Houe M and Townsend P D 1995 Appl. Phys. Lett. 66 2667 [99] Townsend P D 1993 Adv. SolidState Chem. 3 271 [IOO] Nakamura K, Ando H and Shimizu H 1987 Appl. Phys. Lett. 50 1413 [IO11 Carruthers J R 1971 J. Appl. Phys. 42 1846 [IO21 Yamamoto K, Mizuuchi K, Takeshige K, Sasai Y and Taniuchi T 1991 J. Appl. Phys. Lex 70 1947 [IO31 Yamamoto K and Mizuuchi K 1992 IEEE Photon. Tech. Lett. 4 435 [IO41 Fujimura M, Suhara T and Nishihara H 1991 Electron. Lett. 27 1207 [lo51 Miyazawa S 1979 J. Appl. Phys. 50 4599 11061 Peuzin J C I986 Appl. Phys. Lett. 48 1104 [IO71 Miyazawa S 1986 AppL Phys. Lett. 48 1104 [lo81 Thaniyavam S, Finidakly T , Booher D and Moen J 1985 Appl. Phys. Lett. 46 933 [lo91 Lim E J, Fejer M M and Byer R L 1989 Elecrron Lett. 25 174 [110] Lim E J. Fejer M M, Byerand R L and Kozlovsky W J 1989 Electron. Lett. 25 731 [l I I] Cao X,Srivastava R and Ramaswany R VI992 Appl. Phys. Lett. 60 3280 [112] Magel G A, Fejer M M and Byer R L 1990 Appl. Phys. Lett. 56 108 [113] Marowsky G , Canto-Said E J, Lehmann S and Sieverdes F, Bratz A 1993 Phys. Rev. B 48 18 114

1763