Heteroepitaxy, an Amazing Contribution of Crystal Growth to the World of Optics and Electronics

Review Heteroepitaxy, an Amazing Contribution of Crystal Growth to the World of Optics and Electronics

Vladimir L. Tassev

Air Force Research Laboratory, Sensors Directorate, WPAFB, Dayton, OH 45433, USA; [email protected]; Tel.: 1-937-713-8995

Academic Editor: Helmut Cölfen Received: 21 April 2017; Accepted: 11 June 2017; Published: 19 June 2017

Abstract: Advances in Electronics and Optics are often preceded by discoveries in Crystal Growth theory and practice. This article represents in retrospect some of the most significant contributions of heteroepitaxy in these and some other areas—the strong impact of the three modes of heteroepitaxy on microelectronics and quantum optics, the big “push” of PENDEO epitaxy in development of Light Emitting Diodes, etc. A large part of the text is dedicated to heteroepitaxy of nonlinear optical materials grown on orientation-patterned templates and used in the development of new quasi-phase-matching frequency conversion laser sources. By achieving new frequency ranges such sources will result in a wide variety of applications in areas such as defense, security, industry, medicine, and science. Interesting facts from the scientific life of major contributors in the field are mixed in the text with fine details from growth experiments, chemical equations, results from material characterizations and some optical and crystallographic considerations—all these presented in a popular way but without neglecting their scientific importance and depth. The truth is that often heteroepitaxy is not just the better but the only available option. The truth is that delays in device development are usually due to gaps in materials research. In all this, miscommunication between different scientific communities always costs vain efforts, uncertainty, and years of going in a wrong scientific direction. With this article we aim to stimulate a constructive dialog that could lead to solutions of important interdisciplinary scientific and technical issues.

Keywords: heteroepitaxy; hydride vapor phase epitaxy; selective growth; III-Nitrides; nonlinear optical materials; orientation-patterned templates; frequency conversion; quasi-phase-matching

1. Introduction Like many others this story starts at an unexpected time and place. The time is about a year or two before the beginning of World War II and the place is a small ofﬁce in the basement of Iowa State College. John Atanasoff, a young professor with a bachelor degree in materials science, a master’s in mathematics, and a doctorate in theoretical physics, was spending the long winter evenings of 1937–1938 in vain attempts to realize one of his old ideas. Still working on his PhD thesis, Atanasoff had experienced serious computing problems—no matter that his Monroe calculator was one of the best at that time, it was still working so slowly so he had to spend hours and hours in boring calculations. Then, several years later, it was still about the same. The current experiments with vacuum tubes, radio signals and various electronic devices also required time consuming heavy calculations. The idea to help himself by developing his own computing machine without the problems of the existing analog devices—working slowly with accuracy that is strongly dependent upon the performance of each one of their parts—started to obsess him again, more and more, but... with not much success.

Crystals 2017, 7, 178; doi:10.3390/cryst7060178 www.mdpi.com/journal/crystals Crystals 2017, 7, 178 2 of 38

“... One cold winter night, frustrated after so many discouraging events, John got into his Crystalscar and 2017 started, 7, 178 driving without any destination. 200 miles later, he pulled onto a roadside2 of 38 somewhere“... One cold in the winter state night, of Illinois. frustrated Here, after he so continued many discouraging thinking events, about John the creationgot into his of the machinecar and and, started as the driving rumor without says, heany had destination. a drink of200 bourbon. miles later, No he longer pulled nervousonto a roadside and tense, he realizedsomewhere that in his the thoughts state of Illinois. were coming Here, he together continued clearly” thinking [1]. about the creation of the machine and, as the rumor says, he had a drink of bourbon. No longer nervous and tense, In March 1939 after receiving a grant of $650, Atanasoff was ready to embark in the exciting he realized that his thoughts were coming together clearly” [1]. adventure to build an entirely digital machine. He hired Clifford Berry—a bright electrical engineering student, andIn March from 1939 after until receiving 1941 they a workedgrant of on$650, developing Atanasoff theirwas ready Atanasoff-Berry to embark in Computer the exciting (ABC). adventure to build an entirely digital machine. He hired Clifford Berry—a bright electrical ABC had the size of a big desk (Figure1a), weighed 750 pounds and featured rotating drums for engineering student, and from 1939 until 1941 they worked on developing their Atanasoff-Berry memory consisting of hundreds of capacitors. It had also 300 glowing vacuum tubes and a read/write Computer (ABC). ABC had the size of a big desk (Figure 1a), weighed 750 pounds and featured systemrotating that recorded drums for numbers memory by consisti scorchingng of markshundreds on of cards. capacitors. However, It had ABC also had 300 also glowing all major vacuum features of lasttubes generation and a read/write superpower system computers—a that recorded numbers binary by system scorching of arithmetic, marks on cards. separate However, memory ABC and computinghad also functions, all major features regenerative of last memory,generation parallel superpower processing, computers—a electronic binary ampliﬁers, system of clockedarithmetic, control of electronicseparate operations,memory and and computing a modular functions, design. regenerative In 1997, to memory, pay tribute parallel to theprocessing, late inventors, electronic a team of researchers,amplifiers, engineers,clocked control faculty of electronic members, operations, retired faculties and a modular and students design. In from 1997, Iowa to pay State tribute University to and Amesthe late Laboratory inventors, builta team a of working researchers, replica engineer of ABC.s, faculty The project members, took retired four years faculties and and cost students nearly $350. This moneyfrom Iowa compared State University to the $650 and that Ames John Laboratory and Clifford built hada working in 1939 replica (equivalent of ABC. to The only project $8K intook 1997) is four years and cost nearly $350. This money compared to the $650 that John and Clifford had in 1939 an indication of the passion that Atanasoff and Berry had and their hard work. The computers today (equivalent to only $8K in 1997) is an indication of the passion that Atanasoff and Berry had and are, of course, incomparably faster and much more powerful. As for prices and sizes some of them, their hard work. The computers today are, of course, incomparably faster and much more powerful. like RaspberryAs for prices Pi and [2], sizes can besome as of small them, as like the Raspberry credit card Pi that[2], can we be could as small use as to the buy credit them card on-line that we for the ridiculouscould use price to ofbuy . .them . $5 (Figureon-line 1forb). the ridiculous price of … $5 (Figure 1b).

Figure 1. (a) Prof. J. Atanasoff (Iowa State College, 1938) and the replica of ABC (Museum of Figure 1. (a) Prof. J. Atanasoff (Iowa State College, 1938) and the replica of ABC (Museum of Computer Computer History, Mountain View CA) [1]; (b) Raspberry Pi—a credit-card-size computer on the History, Mountain View CA) [1]; (b) Raspberry Pi—a credit-card-size computer on the background of its background of its cost—a five dollar bill (Cambridge UK, 2012) [2]; (c) J. Bardeen, W. Shockley and cost—aW. fiveBrattain—inventors dollar bill (Cambridge of the first UK, semico 2012)nductor [2]; (c) transistor J. Bardeen, (Bell W. Labs, Shockley 1947) and[3]. W. Brattain—inventors of the first semiconductor transistor (Bell Labs, 1947) [3]. What has made these and many other “miracles” possible in the time frame of a human life Whatonly? hasObviously, made these memorable and many events other like “miracles” the demonstration possible inin the1947 time of framethe first of semiconductor a human life only? Obviously,transistor memorable (Figure 1c). events However, like a the closer demonstration look under the in ordinary 1947 of theengineering first semiconductor surface will reveal transistor dozens of examples of great discoveries in crystal growth theory and practice that are the real reason (Figure1c). However, a closer look under the ordinary engineering surface will reveal dozens of for the rapid progress in the last a hundred years or so—a period that starts in the dawn of the examples of great discoveries in crystal growth theory and practice that are the real reason for the crystal growth theory. The following paragraphs are dedicated to some of them. rapid progress in the last a hundred years or so—a period that starts in the dawn of the crystal growth theory.2. The The Growth following Modes paragraphs of Heteroepitaxy are dedicated to some of them.

2. The GrowthIn 1939, Modes about ofthe Heteroepitaxy same time when Atanasoff and Berry were working on their ABC, six thousand miles east of Iowa State College somewhere on the Balkans, another ambitious professor Infrom 1939, the about Department the same of Physical time when Chemistry Atanasoff of andSofia Berry University, were workingIvan Stranski on their (Ivan ABC, is the six Slavic thousand milestranslation east of Iowa of John) State and College his brilliant somewhere former on student the Balkans, and co-worker another ambitiousnow, Lubomir professor Krastanov, from the Departmentcontinued of to Physical be interested Chemistry in heteroepitaxy. of SoﬁaUniversity, They were well-grounded Ivan Stranski with (Ivan what isthe had Slavic already translation been of John) and his brilliant former student and co-worker now, Lubomir Krastanov, continued to be Crystals 2017, 7, 178 3 of 38 interested in heteroepitaxy. They were well-grounded with what had already been done in the ﬁeld so they had a great deal of thoughts about the two well-known cases of heteroepitaxy—the two-dimensional (2D) layer-by-layer growth—the so-called Frank–Van der Merwe (FM) mode, and the three-dimensional (3D) island growth, known as Volmer–Weber (VW) mode. However, they felt that something was missing. For example, they wondered what pulls the trigger of converting one smooth at the beginning 2D layer-by-layer (FM) growth in a rough at the end 3D island (VW) growth. They believed that the so-called “half-crystal position”, introduced by them, called “kink” now, played somehow an important role in explanation of these processes. The unicity of this position is that one can increase or decrease the crystal size simply by successive, reversible attachments or detachments of atoms at the kink site, and no change in the surface energy will take place. At the same time the energy released at an act of attachment will be exactly equal to three times the single work of separation, i.e., 3ψ, as far as three neighboring units are always involved in this case—one from the raw, one from the plane, and one from the crystal. The work of separation from this position ϕ1/2 is also equal to the lattice energy of the crystal per building unit and taken with opposite sign chemical potential µ, i.e.:

ϕ1/2 = 3ψ = −µ (1)

In the case of heteroepitaxy, i.e., when a crystal material (marked with a “prim”) grows on another, I substrate material, there are two possibilities: (i) when ϕ1/2 > ϕ1/2, then the gradient of the chemical dµ potential in the growth direction is positive, i.e., dn > 0 and the growth of the “prim” material, being energetically favored, will result in a smooth 2D-type surface morphology (Figure2a); and (ii) when I dµ ϕ1/2 < ϕ1/2 and dn < 0, then the growth will be rather 3D-type island growth (Figure2b). To obtain more information Stranski and Krastanov tried to look at the half-crystal position from a different angle. When a foreign (unlike or “prim”) atom attaches to the substrate surface this atom will be connected in a vertical (normal) and a horizontal (lateral) direction. This arrangement allows I the energy released after the attachment at the half-crystal position ϕ1/2 to be expressed differently, as:

I ϕ1/2 = ψlateral + ψnormal (2)

Here ψlateral and ψnormal denote the lateral and the normal energies released after bonding the approaching atom with the half atomic row, the half crystal plane and the half underlying crystal block. This expression is more informative because now we replace the underlying crystal block by another block from a different crystal material. Thus, the lateral bonding with the two neighbors from the row and from the plane will remain about the same as in the case of homoepitaxy. However, the normal I bonding (the one across the interface) will be changed and the changed total work of separation ϕ1/2 at this single act of attachment will be:

I I ϕ1/2 = 2ψ + ψ (3)

Now by adding and extracting the single work of separation ψ, Equation (3) converts in:

I I I I ϕ1/2 = 2ψ + ψ = 3ψ − (ψ − ψ ) = ϕ1/2 − (ψ − ψ ) (4)

Here ψI is the single work of separation of unlike (foreign, or “prim”) particles, while ψ is the single work of separation for identical (own) particles. By having (ψ − ψI ) in the formula, it is easily to I I I say that, if ψ < ψ , then ϕ1/2 > ϕ1/2 and the equilibrium vapor pressure of the ﬁrst monolayer p∞(1) on the foreign substrate will be smaller than the equilibrium vapor pressure of the bulk crystal p∞, i.e., I p∞(1) < p∞. Then at least one monolayer can be deposited at any vapor pressure p that is higher than I p∞(1) even at under saturation with respect to the bulk crystal. In the opposite case, the opposite is true, and a higher supersaturation is required [4]. Crystals 2017, 7, 178 4 of 38

ψ < ψ I ϕ I > ϕ easily to say that, if , then 1 / 2 1 / 2 and the equilibrium vapor pressure of the first I monolayer p∞ (1) on the foreign substrate will be smaller than the equilibrium vapor pressure of I the bulk crystal p∞ , i.e., p∞ (1) < p∞ . Then at least one monolayer can be deposited at any I vapor pressure p that is higher than p∞ (1) even at under saturation with respect to the bulk Crystals 2017, 7, 178 4 of 38 crystal. In the opposite case, the opposite is true, and a higher supersaturation is required [4].

FigureFigure 2. 2.The The three three major major modes modes of of heteroepitaxy: heteroepitaxy: (a ()a the) the 2D 2D layer-by-layer layer-by-layer growth growth called called Frank–Van Frank–Van derder Merwe Merwe (FM) (FM) mode; mode; (b ()b the) the 3D 3D island island growth growth called called Volmer–Weber Volmer–Weber (VW) (VW) mode; mode; and and (c )(c the) the layer-plus-islandlayer-plus-island growth growth called called Stranski–Krastanov Stranski–Krastanov (SK) (SK) mode. mode.

DevelopingDeveloping further further this this idea, idea, Stranski Stranski and and Krastanov Krastanov concluded concluded that that after after the the 2D 2D layer-by-layer layer-by-layer growthgrowth of severalof several monolayers, monolayers, at a certainat a certain critical critical thickness thicknesshc (which hc depends (which ondepends the lattice on mismatchthe lattice betweenmismatch the between film and the film substrate), and the 3D substrate), islands will 3D start islands to appear will start on to the appear film surface on the and film the surface growth and willthe continue growth three-dimensionally.will continue three-dimensionally. The same idea canThe be same equivalently idea can formulated be equivalently as island formulated formation as onisland the top formation of a wetting on the layer top with of a thickness-dependent wetting layer with surfacethickness-dependent energy. This approach surface energy. allows theThis critical wet layer thickness hc to be ab initio calculated [5,6]. The formation of the 3D islands is approach allows the critical wet layer thickness hc to be ab initio calculated [5,6]. The formation of accompanied with the formation of some edge dislocations within them, created in response to the the 3D islands is accompanied with the formation of some edge dislocations within them, created in need to reduce the strain between the substrate and the film [7]. This intermediate Stranski–Krastanov response to the need to reduce the strain between the substrate and the film [7]. This intermediate (SK) “layer-plus-island growth” (Figure2c), is one of the three known modes of heteroepitaxy classified Stranski–Krastanov (SK) “layer-plus-island growth” (Figure 2c), is one of the three known modes of later, in 1958, by Ernst Bauer [8]. heteroepitaxy classified later, in 1958, by Ernst Bauer [8]. Meantime scientists and engineers put them into practice. From this point there is only a little Meantime scientists and engineers put them into practice. From this point there is only a little step to the invention of the first transistor that soon will replace all 300 glowing vacuum tubes in ABC step to the invention of the first transistor that soon will replace all 300 glowing vacuum tubes in and “shrink” the Electronics to Microelectronics. The three modes of heteroepitaxy, however, are of ABC and “shrink” the Electronics to Microelectronics. The three modes of heteroepitaxy, however, great importance not only for Electronics but for Optics as well. Today the SK mode is widely used are of great importance not only for Electronics but for Optics as well. Today the SK mode is widely in the theory and practices of the formation of quantum dots (Figure3a) [ 9] and the FM mode—in used in the theory and practices of the formation of quantum dots (Figure 3a) [9] and the FM the quantum wells (Figure3b) [ 10]. A typical example is [11] where the authors evaluate the wetting mode—in the quantum wells (Figure 3b) [10]. A typical example is [11] where the authors evaluate layer and its temperature dependence in order to determine the optimal conditions for self-assembly the wetting layer and its temperature dependence in order to determine the optimal conditions for of InAs/GaAs quantum dots. self-assembly of InAs/GaAs quantum dots.

Crystals 2017, 7, 178 5 of 38 Crystals 2017, 7, 178 5 of 38

FigureFigure 3. TransmissionTransmission electron electron micr microscopyoscopy (TEM) (TEM) images images of: of: ( (aa)) InGaN InGaN quantum quantum dots dots [9] [9] formed formed withinwithin a a GaN structure and ( b)) AlGaN/AlN quantum quantum wells wells [10], [10], both both within within an an AlN AlN structure. structure. (Reproduced(Reproduced from from [9,10] [9,10] with with the the permission of AIP Publishing).

3.3. III-Nitrides III-Nitrides

3.1.3.1. Heteroepitaxy Heteroepitaxy of of III-Nitrides III-Nitrides III-Nitrides,III-Nitrides, such such as as GaN GaN and and AlN, AlN, due due to to their their wideband wideband stru structure,cture, high high thermal thermal conductivity, conductivity, highhigh melting point,point, andand highhigh resistivity resistivity to to chemicals chemicals are are among among the the most most preferable preferable materials materials for high for highpower, power, high frequency,high frequency, and high and temperature high temperatur electronics;e electronics; and for and bright, for shorterbright, wavelengthshorter wavelength emitters emittersin the LED in the technology. LED technology. These materials, These materials, however, however, cannot be cannot grown be from grown melt from because melt of because their high of theirvapor high pressure vapor and pressure do not and have do their not ownhave templates their own so templates the only optionso the foronly them option is heteroepitaxy. for them is heteroepitaxy.There are a number There of are articles a number proposing of articles different propos traditionaling different [12] or traditional more exotic [12] [13 or] substratemore exotic options. [13] substrateThe most proposedoptions. The substrates most forprop GaNosed are substrates sapphire, SiC,for GaN and AlN,are butsapphire, the problem SiC, and with AlN, the last but one the is problemthe same—AlN with the itself last does one notis the have same—AlN its own templates. itself does SiC not is have expensive its own and templates. has quite SiC different is expensive thermal andexpansion, has quite which different means thermal “expect expansion, troubles” which especially means at thick“expect layer troubles” growth. especially The cheap at andthick readily layer growth.available The sapphire cheap seemsand readily to be theavailable best heteroepitaxial sapphire seems option to be forthe thin best or heteroepitaxial thick GaN growth. option However, for thin orthe thick lattice GaN mismatch growth. between However, GaN the and lattice sapphire mismatch is huge, between 33% (Figure GaN4 ),and which sapphire makes is “the huge, mission 33% (Figurealmost impossible”.4), which makes One elegant“the mission solution, almost which impo for assible”. long time One now elegant is a practice solution, in thewhich semiconductor for a long timeindustry, now isis toa practice nitride ﬁrst in the the semiconductor sapphire surface indust withry, ammonia is to nitride and first after the that sapphire to deposit surface a thin with low ammoniatemperature and GaN after buffer that to layer. deposit While a thin the bufferlow temp layererature aims toGaN reduce buffer the layer. stress While between the twobuffer different layer aimsmaterials, to reduce the purposethe stress of between the nitridation two different is to formmaterials, a thin the AlN purpose layer priorof the to nitridation the deposition is to form of the a thinGaN AlN layer: layer prior to the deposition of the GaN layer: 3 NH3 + Al2O3 = AlN + H2O2 (5) + = + 32 NH3 Al2O3 AlN H 2O2 (5) The expectation is that due to the smaller lattice mismatch2 between AlN and GaN, 2.5%, GaN can beThe grown expectation more easilyis that on due AlN to thanthe smaller on sapphire. lattice However,mismatch abetween new problem AlN and arises. GaN, The 2.5%, in-plane GaN canlattice be mismatchgrown more between easily AlNon AlN (0001) than and on sapphire sapphire. (0001) However, is also a large, new evenproblem larger, arises. 35%. The Fortunately, in-plane ◦ latticewhen AlNmismatch (0001) between grows on AlN sapphire (0001) (0001),and sapphire it naturally (0001) rotates is also at large, 30 (Figure even larger,4) around 35%. its Fortunately, c-axis with whenrespect AlN to sapphire.(0001) grows This on rotation sapphire reduces (0001), the it AlN/sapphire naturally rotates lattice at 30° mismatch (Figure from 4) around 33% to its 13.3% c-axis so withAlN respect grows on to sapphire.sapphire. This rotation reduces the AlN/sapphire lattice mismatch from 33% to 13.3% so AlN grows on sapphire.

CrystalsCrystals 20172017, ,77, ,178 178 66 of of 38 38

Figure 4. Top view of the epitaxial relationship between AlN (0001) and sapphire (0001) with 30◦ Figurerotation 4. withTop respectview of to the each epitaxial other. The relationship distance marked between as AlN “a” is (0001) the separation and sapphire between (0001) O2− withatoms 30° in rotationsapphire with [14]. respect (Reproduced to each fromother. [14 The] with distance some marked modifications). as “a” is the separation between O2− atoms in sapphire [14]. (Reproduced from [14] with some modifications). Metal Organic Chemical Vapor Deposition (MOCVD) growths of high quality GaN using an AlN bufferMetal layer Organic were performed Chemical still Vapor in the Deposition mid-80s by (MOCVD) Amano [15 growths] who also of realized high quality growth GaN of AlGaN using andan AlNGaInN buffer ternary layer alloys were [16performed].Growth still of GaN in the using mid-80 a lows by temperature Amano [15] GaN who buffer also layer realized was growth performed of AlGaNfirst by and Nakamura GaInN [ternary17] who alloys also reported [16].Growth the highest of GaN at using the time a low Hall temperature mobility of GaN.GaN buffer layer was performed first by Nakamura [17] who also reported the highest at the time Hall mobility of GaN.3.2. Nano- and Micro-Heteroepitaxy Instead of using a low temperature buffer layer, there are other ways to relieve the strain between 3.2. Nano- and Micro-Heteroepitaxy the sapphire substrate and the growing GaN layer, for example, by the so-called AGOG (Aluminum, Growth,Instead Oxide, of using Grain) a approachlow temperature [18,19]. Thebuffer AGOG layer, approach there are consists other ofways the nextto relieve steps: the (i) coatingstrain betweenthe sapphire the sapphire surface with substrate a thin and Al layer; the growing (ii) making GaN a pattern—aslayer, for example, shown on by Figure the so-called5a the pattern AGOG is (Aluminum,composed of Growth, many 400 Oxide, nm × Grain)400 nm approach× 100 nm [18,19]. miniature The square AGOG mesas; approach (iii) treating consists the of film, thefirst next at ◦ steps:lower (i) temperature coating the (450sapphireC) to surface oxidize with it, i.e., a thin to turn Al layer; the mesas (ii) making into polycrystalline a pattern—as Alshown2O3 (Figure on Figure5b), 5aand the after pattern that is at composed higher temperature of many 400 (1250 nm ◦×C) 400 to nm convert × 100 the nm mesas miniature into crystallinesquare mesas; form (iii) (Figure treating5c). theThe film, AGOG first technique at lower enablestemperature the hard (450 steps °C) to of polishingoxidize it,and i.e., patterningto turn the the mesas resistant into sapphirepolycrystalline surface Alto2O be3 (Figure avoided, 5b), or toand nitride after that it. Moreover, at higher thetemperatur sapphiree islands(1250 °C) act to as convert springs the that mesas can reduce into crystalline the strain formand eventually(Figure 5c). facilitate The AGOG the separation technique of enables the film fromthe hard the substrate.steps of polishing and patterning the resistantStrain sapphire induced surface by different to be avoided, lattice parameters or to nitride and it. different Moreover, thermal the sapphire properties islands may offeract as an springsadditional that degree can reduce of freedom the strain for tailoring and eventually materials, facilitate but this the isoften separation at the expenseof the film ofdislocation from the substrate.generation and even cracking of the growing layer. Recently, eliminating some of the drawbacks, a low temperatureStrain induced 3D vertical by different epitaxial lattice growth parameters of micrometer-scale and different defect-free thermal Ge properties and SiGe may crystals offer on an Si additionalwas demonstrated degree of [ 20freedom], indicating for tailoring that similar materials, advantages but this can is be often achieved at the without expense using of dislocation patterned generationsubstrates. and Applying even cracking this innovative of the growing strategy, layer. when Recently, exploiting eliminating strongly out-of-equilibriumsome of the drawbacks, growth a lowconditions, temperature defect-free 3D vertical mismatched epitaxial hetero-structures growth of micrometer-scale on Si substrates defect-free were also Ge produced and SiGe [21 crystals]. In this oncase Si thewas strain demonstrated relaxation [20], was indicating engineered that to occursimilar elastically advantages rather can than be achieved plastically without by combining using patternedsuitable substrate substrates. patterning Applying and vertical this crystal innovative growth withstrategy, compositional when grading. exploiting All this strongly extended out-of-equilibrium growth conditions, defect-free mismatched hetero-structures on Si substrates were also produced [21]. In this case the strain relaxation was engineered to occur elastically rather

Crystals 2017,, 7,, 178178 7 of 38

than plastically by combining suitable substrate patterning and vertical crystal growth with thecompositional concept from grading. nano to All the this micron extended scale heteroepitaxy.the concept from Although nano to for the now micron the concept scale heteroepitaxy. is proven only forAlthough the case for of now SiGe/Si, the concept the predictions is proven are only that for the the idea case will of be,SiGe/Si, eventually, the predictions extended are to heteroepitaxy that the idea ofwill many be, eventually, other dislocation-free extended to materials heteroepitaxy [22]. of many other dislocation-free materials [22].

Figure 5.5. Scanning electron microscopymicroscopy (SEM) images of: ( a) the patterned by e-beame-beam lithographylithography sapphire surfacesurface covered covered with with ordered ordered metallic metallic aluminum aluminum mesas. mesas. The The average average sizes sizes of the of square the square mesas aremesas 400 are nm 400× 400nm nm× 400× nm100 × nm; 100 ( bnm;) the (b pattern) the pattern after lowafter temperature low temperature annealing—the annealing—the islands islands have

convertedhave converted into polycrystalline into polycrystalline Al2O3; and Al2 (Oc)3; the and pattern (c) the after pattern the solid after state the crystalline solid state conversion crystalline [19]. (Reproducedconversion [19]. from (Reproduced [19] with the from permission [19] with of theJ. Mat.permission Res.). of J. Mat. Res.).

3.3. Combinations of Different Growth Techniques As it was shown in Section 3.1[ [15,16],15,16], the formationformation of AlN on the sapphire substrate by nitridation cancan bebe replacedreplaced byby a a preliminary preliminary deposition deposition of of a 1–2a 1–2µm µm thin thin AlN AlN layer. layer. In someIn some cases, cases, for examplefor example when when SiC isSiC used is asused a substrate, as a substrate, this is actuallythis is actually the only the available only available option for option thickgrowth for thick of III-Nitrides.growth of III-Nitrides. Due to the Due high to surface the high quality surface of the qual sapphireity of the or sapphire SiC wafers, or theSiC depositionwafers, the of deposition this layer shouldof this layer be done should in a be growth done processin a growth that isprocess far from that thermodynamic is far from thermodynamic equilibrium, equilibrium, i.e., a growth i.e., that a doesgrowth not that need does surface not need defects surface to initiate defects nucleation, to initiate a nucleation, technique that a technique provides that conditions provides for conditions rather 2D thanfor rather for 3D 2D growth. than for Such 3D techniques growth. Such are, fortechniqu example,es are, the Molecularfor example, Beam the Epitaxy Molecular (MBE) Beam or the Epitaxy Metal Organic(MBE) or Chemical the Metal Vapor Organic Deposition Chemical (MOCVD). Vapor However, Deposition although (MOCVD). these techniquesHowever, arealthough suitable these for a thintechniques layer growth, are suitable they arefor completelya thin layer incapablegrowth, they of achieving are completely thick, forincapable example, of hundredsachieving ofthick, micron for layersexample, that hundreds are required of micron for many layers electronic that are and required optical applications.for many electronic Then, the and second optical growth applications. step on theThen, thin the MBE second or MOCVD growth layer step muston the be thin performed MBE or by MOCVD another growthlayer must technique be performed capable of by producing another thisgrowth thick technique growth. Thecapable Halide of producing Vapor Phase this Epitaxy thick growth. (HVPE) The is one Halide of these. Vapor Combining Phase Epitaxy two and (HVPE) even moreis one different of these. growth Combining techniques two and to achieveeven more a device different structure growth is nottechniques an unusual to achieve approach a device in the semiconductorstructure is not III-Nitridean unusual industry. approach More in the details semiconductor related to otherIII-Nitride wideband indust semiconductorry. More details materials related (GaAsto other and wideband GaP) are semiconductor further presented materials in Section (GaAs 6.5 and. GaP) are further presented in Section 6.5.

3.4. Crystal Defects in III-Nitrides and Other III-V Compound Semiconductors The development of semiconductor light sources and, particularly, of Light Emitter Diodes (LED) is an indivisible part of our World whichwhich isis directlydirectly relatedrelated toto cleanclean environmentenvironment and energy conversion. InvolvingInvolving more more and and more more wideband wideband semiconductor semiconductor materials materials made itmade possible it possible to produce to highproduce intensity high intensity sources thatsources radiate that inradiate red, yellow,in red, yellow, and green. and However,green. However, blue was blue missing was missing for years for fromyears thefrom LED the spectrum LED spectrum and, from and, here, from the here, white the aswh aite mixture as a mixture of red, green,of red, and green, blue. and Because blue. Because of their widebandof their wideband structure, structure, everybody everybody was waiting was for wait improvementsing for improvements in the III-Nitride in the growth III-Nitride technologies. growth Intechnologies. the case of In GaAs, the case GaP of and Ga otherAs, GaP related and traditionalother related III-V traditio semiconductornal III-V semiconductor materials, the materials, efficiency ofthe a efficiency LED is strongly of a LED dependent is strongly on dependent the dislocation on the density dislocation in the density material in [23 the] (Figure material6a). [23] Based (Figure on this6a). experience,Based on this it looked experience, almost it impossible looked almost that GaN, impossible due to itsthat high GaN, dislocation due to density—sometimesits high dislocation 10 2 7 2 ~2.10density—sometimes/cm [24], but ~2.10 definitely10/cm2 always [24], but >10 definitely/cm —could always be used >107 in/cm making2—could efficient be used optoelectronic in making devicesefficient (Figureoptoelectronic6b). devices (Figure 6b). Fortunately, some profound studies [25] showed that the performance of the optoelectronic devices based on nitride semiconductors is not that sensitive to high concentration of extended

Crystals 2017, 7, 178 8 of 38

Fortunately, some profound studies [25] showed that the performance of the optoelectronic devices Crystals 2017, 7, 178 8 of 38 based on nitride semiconductors is not that sensitive to high concentration of extended defects as it is indefects the case as it of is the in traditionalthe case of III-V the traditional compounds. III-V As acompounds. matter of fact, As these a matter two familiesof fact, these are quite two different.families First,are quite the traditionaldifferent. First, III-V compoundsthe traditional are mostlyIII-V compounds covalent, while are mostly the chemical covalent, bonds while in the the III-Nitrides chemical arebonds strongly in the ionic.III-Nitrides This leadsare strongly to the bunchingionic. This oflead thes to surface the bunching states as of well the assurface the states states associated as well as withthe states dangling associated bonds with in edge dangling dislocations bonds in near edge the dislocations band edges, near which the band prevents edges, them which from prevents being non-radiativethem from being recombination non-radiative centers. recombination Second, the cent III-Nitridesers. Second, can existthe III-Nitrides in both equilibrium can exist wurtzitein both andequilibrium metastable wurtzite cubic structureand metastable as the enthalpy cubic struct of formationure as the of theseenthalpy two allotropicof formation forms of differs these onlytwo byallotropic a few meV. forms Thus differs conversions only by between a few meV. the two Thus phases conversions can occur between easily by the creation two phases of stacking can faultsoccur alongeasily theby closed-packedcreation of stacking (0001) faults and (111) along planes. the closed-packed As a result, basal (0001) plane and stacking (111) planes. faults are As abundant a result, inbasal both plane compounds stacking and faults alloys. are abundant This leads in to both strong compounds band structure and alloys. potential This fluctuations. leads to strong Finally, band in somestructure of the potential III-Nitride fluctuations. alloys, such Finally, as InGaN in some and of AlGaN, the III-Nitride some additional alloys, such band as InGaN structure and potential AlGaN, fluctuationssome additional also band exist, structure due to phase potential separation fluctuations and alloyalso exist, ordering. due to These phase potential separation fluctuations and alloy contributeordering. These additionally potential to excitonfluctuatio localizationns contribute and additionally thus allow efficient to exciton radiative localization recombination and thus even allow at roomefficient temperature. radiative recombination even at room temperature.

Figure 6. ((a)) TheThe normalizednormalized LightLight Emitting Diodes (LED) efﬁciencyefficiency as a function of thethe dislocationdislocation density in some traditional III-V semiconductor compounds. [23] [23] (Reproduced from [23] [23] with the permission ofof AIPAIP Publishing);Publishing); ((bb)) cross cross sectional sectional view view of of a a Nichia Nichia GaN/GaInN/GaN/Sapphire GaN/GaInN/GaN/Sapphire based LED [24[24]] with with dislocation dislocation density density in the in range the ofrange 1010 /cmof 102.10 (Reproduced/cm2. (Reproduced from [24 ]from with the[24] permission with the ofpermissionNature). of Nature).

3.5. PENDEO and ELO Epitaxy Work [25] is a classic example from the world of science and technology of how things that wereWork considered [25] is for a classic years exampleas impossible from thecould world be made of science possible. and This technology temporary of how relief, things however, that were did considerednot stop the for attempts years as to impossible improve couldthe III-Nitride be made possible.material Thisquality temporary and explain relief, the however, complex did nature not stop of thethis attemptstypical “under-compulsion” to improve the III-Nitride heteroepitaxial material quality growth and [26]. explain In all thethis, complex the idea nature of using of this patterned typical “under-compulsion”substrates was also not heteroepitaxial neglected. Thus growth enhanced [26]. In alllight this, emitting the idea diodes of using (LED) patterned with relatively substrates high was alsooutput not power neglected. were Thus fabricated enhanced from light InGaN emitting [27] diodes and (LED)GaN [28] with on relatively sapphire high patterned output powersubstrates. were fabricatedHowever, fromreal progress InGaN [27 in] andLED GaN technology [28] on sapphire was achieved patterned after substrates. discoveries However, of techniques real progress such as in LEDEpitaxial technology Lateral wasOvergrowth achieved (ELO) after discoveries[29,30] and ofPENDEO techniques (from such Latin: as Epitaxial to hang Lateralor to be Overgrowthsuspended) (ELO)epitaxy [29 [30,31].,30] and In PENDEO these cases (from the Latin: deposition to hang of ora buffer to be suspended) layer is worthwhile epitaxy [30 because,31]. In theseit absorbs cases the depositionstrain built ofin a the buffer growing layer is material worthwhile and becauseblocks some it absorbs of the the treading strain built dislocations. in the growing Because material it is andenergetically blocks some more of thefavorable treading the dislocations. propagating Because dislocatio it isns energetically to follow the more front favorable of crystallization the propagating (being perpendicular to it), a switch in the growth direction from vertical to lateral would result in strong reduction of the dislocations that otherwise appear on the top layer surface—similarly to this case, the dislocation propagation during Czochralski (CZ) melt growth can be reduced by the so-called

Crystals 2017, 7, 178 9 of 38 dislocations to follow the front of crystallization (being perpendicular to it), a switch in the growth direction from vertical to lateral would result in strong reduction of the dislocations that otherwise appear on the top layer surface—similarly to this case, the dislocation propagation during Czochralski Crystals 2017, 7, 178 9 of 38 (CZ) melt growth can be reduced by the so-called “necking” the seed. In this process only those dislocations“necking” the that seed. are strictlyIn this inprocess the growth only those direction disloc mayations propagate that are through strictly the in neck,the growth while alldirection others willmay bepropagate restricted. through the neck, while all others will be restricted. To ensureensure the the change change in thein the crystallization crystallization direction direction during during ELO or ELO PENDEO or PENDEO epitaxy, theepitaxy, substrate, the whichsubstrate, is already which coveredis already with covered a thin with SiO2 ora thin Si3N SiO4 layer2 or thatSi3N prohibits4 layer that any prohibits growth, isany patterned growth, byis photolithography.patterned by photolithography. By etching the By prohibiting etching the layer proh downibiting to thelayer substrate, down to we the expose substrate, periodical we expose areas openperiodical for growth. areas open These for “windows”, growth. These usually “windows”, in the form usually of narrow in theparallel form ofstrips, narrow are parallel separated strips, by muchare separated wider “streets”, by much where wider growth“streets”, is notwher welcomede growth (Figureis not welcomed7a). (Figure 7a).

Figure 7. ((aa)) Different Different stages stages of of an an Epitaxial Epitaxial Latera Laterall Overgrowth (ELO) growth—the front of crystallization may may change change or or not not its its direction; direction; (b (b) )PENDEO PENDEO epitaxy: epitaxy: the the presence presence of ofa buffer a buffer layer layer is isnot not necessary; necessary; in inboth both cases cases blocking blocking the the vertical vertical growth growth on onthe theridges ridges may may be purposely be purposely done done by the by thedeposition deposition of a oflayer a layer that that prohibits prohibits the thegrowth growth [31]. [31 (Reproduced]. (Reproduced from from [31] [31 with] with the the permission permission ofof J. J.Cryst. Cryst. Growth Growth). ).

By applyingapplying initially initially conditions conditions for for a vertical a vertical growth growth the growth the growth is favored is favored through thethrough openings. the Asopenings. a result, As ridges a result, of crystalline ridges of materialcrystalline are material formed inare the formed vertical in direction. the vertical After direction. the ranges After “gain” the enoughranges “gain” height enough above the height “prohibiting” above the layer “prohibiting” surface, the layer growth surface, conditions the growth should conditions be changed should in such be achanged way so asin tosuch promote a way now so moreas to lateral promote than now vertical mo growth.re lateral This than growth vertical continues growth. until This the growth ridges touchcontinues each until other the covering ridges thetouch areas each between other them,covering depending the areas on between the growth them, conditions, depending the growthon the overgrowth the conditions, prohibited the areas growth may evenover the not prohibited touch the surfaces,areas may forming even not voids touch underneath the surfaces, (Figure forming7b). Thus,voids verticallyunderneath propagating (Figure 7b). dislocations Thus, vertically will appear propagating only on dislocations the top surface will above appear the only narrow on the ridges, top wheresurface the above growth the narrow is still in ridges, a vertical where direction. the growth In another is still scenarioin a vertical the direction. narrow vertical In another ridges scenario can be formedthe narrow chemically vertical byridges etching can betweenbe formed the chemical wide streetsly by to etching a depth between that is aboutthe wide equal streets to the to expected a depth thicknessthat is about of theequal growing to the layer.expected In such thickness a case of growth the growing will be layer. only lateralIn such and a case it will growth be not will necessary be only tolateral change and theit will direction be not necessary of crystallization to change from the verticaldirection to of lateral. crystallization In other from cases vertical the top to surfacelateral. In of theother vertical cases ridgesthe top could surface be of covered the vertical with anotherridges could prohibiting be covered layer with in order another additionally prohibiting to suppress layer in theorder vertical additionally and promote to suppress the lateral the growth.vertical Thisand steppromote is almost the lateral mandatory growth. in theThis case step of thinis almost layer growth,mandatory when in the thickness case of controlthin layer is difﬁcult. growth, Thiswhen is thickness because whencontrol the is thicknessdifficult. This of the is growingbecause when layer decreasesthe thickness the of radius the growing of the curvature layer decreases of its side the wall radius decreases, of the curvature as well. Thisof its leads side wall to a localdecreases, increase as ofwell. the This equilibrium leads to a concentration local increase in of the the vapor equilibrium phase aroundconcentration this wall in duethe vapor to the phase Gibbs-Thompson around this effect.wall due In to this the case Gibbs-Thompson the supersaturation effect. is In high this enough case the to supersaturation promote two-dimension is high enough nucleation to promote of the layer,two-dimension but it can nucleation also promote of the vertical layer, growth.but it can One also should promote bear vertical in mind growth. that there One isshould a signiﬁcant bear in differencemind that betweenthere is “buffer”a significant and “prohibiting” difference betw layer.een While “buffer” the “buffer” and “prohibiting” layer aims to layer. reduce While the stress the “buffer” layer aims to reduce the stress and dislocation density accommodating two different materials to each other during heteroepitaxy, the “prohibiting” layer is just a layer that stops the growth. A prohibiting layer, if necessary, can be deposited on an already deposited buffer layer. Buffer layers can be used to “filtrate” the propagating dislocations in the case of homoepitaxy as well, as far as the interface, as a highly defective area is a “generator” of dislocations itself. Thus, for

Crystals 2017, 7, 178 10 of 38 and dislocation density accommodating two different materials to each other during heteroepitaxy, the “prohibiting” layer is just a layer that stops the growth. A prohibiting layer, if necessary, can be deposited on an already deposited buffer layer. Buffer layers can be used to “ﬁltrate” the propagating dislocations in the case of homoepitaxy as well, as far as the interface, as a highly defective area is a “generator” of dislocations itself. Thus, for example, a better crystalline quality Si can be achieved by patterning of the Si-substrate using a SiO2 mask and performing a subsequent lateral over growth.

4. Other Matching Techniques Engineering the buffer layer, i.e., making the right choice of a buffer material that will allow growth of one material on another at a large lattice mismatch can lead to unexpectedly good results. For example, Detchprohm et al. [32] demonstrated that several hundred of microns thick GaN can be grown at a relatively high growth rate of 50–70 µm/h on a ZnO buffer layer deposited on a GaAs substrate although the direct growth of GaN on GaAs with the same technique (HVPE) always results in a polycrystalline GaN. However, there are other approaches that may be also helpful without the deposition of a buffer layer. For example, the lattice constant of the growing layer in some cases could have a better match with the lattice parameter of the substrate when its crystallographic orientation is different. Thus, according to [33] cubic GaN layers can be successfully grown on (100) GaAs substrates while hexagonal GaN fits better to (111) oriented GaAs substrates. However, matching two crystal lattices is only half of the story. Two materials could have an almost perfect crystallographic match but completely different thermal properties, i.e., different thermal conductivity and thermal expansion coefficients. This could be another long line of problems to resolve, especially when the goal is a thick layer growth. The suitable choice of a substrate, crystallographic match, and growth technique (or a combination of growth techniques) plus clever engineering of the buffer layer can make successful many heteroepitaxial combinations that at first sight look impossible.

5. Triumphs and Defeats of the Direct Laser Sources High power, tunable laser sources in the mid- and longwave IR and THz regions are in great demand for a wide variety of military and commercial applications in areas such as defense (aircraft protection, laser radar, and IR communications), security (airport scanners, remote sensing of chemicals, incl. explosives, and biological agents), industry (gas sensing, leak detection, pollution monitoring, process control, etc.), science (IR and THz spectroscopy) and medicine (medical imaging, biopsy-free cancer detection, etc.). However, despite the huge number of pulsed and continuous laser sources developed [34] since 1960 when the ﬁrst Ruby laser was demonstrated at the Hughes Research Labs [35] (Figure8a), there are only a few available direct sources in the IR and THz region. They are the quantum cascade lasers (QCLs) and a few lasers based on ternary or quaternary lead salt compounds such as PbxSn1−xTe and PbxEu1−xSeyTe1−y [36] (Figure8b). These sources are the only ones that cover both atmospheric windows of transparency between 2–5 and 8–12 µm, which are the only wavelength regions where one can detect an optical event or distribute in the atmosphere an IR signal. Relying on fairly small bandgap transitions and thus producing modest output power, however, these sources are, in addition, hardly tunable and usually require cryogenic temperatures to operate properly, because their performance strongly deteriorates with increase in temperature. Fortunately, the inability to achieve the needed wavelengths, power, and tunability using direct laser sources was happily compensated by discoveries related to exploring nonlinear frequency conversion processes such as second harmonic generation (SHG), birefringent, and quasi-phase-matching (QPM), etc. This eventually led to the development of coherent laser sources based on three and four wave mixing within a nonlinear medium. Thus dreamt of wavelength ranges were reached, which resulted in laser applications that went way beyond any initial expectations. New needs and new customer requirements, however, that pop up on a daily basis are a normal thing for our fast changing world. All this brings to scientists and engineers the bad feeling that the mission is not accomplished, yet. Crystals 2017, 7, 178 11 of 38 Crystals 2017, 7, 178 11 of 38

Figure 8.8. (a(a)) Theodore Theodore Maiman Maiman with with the the ﬁrst first Ruby Ruby laser demonstratedlaser demonstrated in 1960 in at 1960 the Hughes at the ResearchHughes ResearchLabs [37]; Labs (b) Some [37]; of(b) the Some available of the directavailable laser direct sources laser in sources the IR and in the THz IR regionand THz [38 region]. [38].

6. Engineering of Materials for Frequency Conversion in the IR and THz Region

6.1. Material Aspects in Retrospect Converting the frequency of an available pump laser into a new wavelength of interest was initially realized in bulk nonlinear birefringent crystals such as AgGaSe2,, ZnGeP ZnGeP22 [39],[39], and and KTP. KTP. However, somesome problems problems with with the the birefringent birefringent crystals crystals such assuch thermal as therma lensing,l lensing, low damage low thresholddamage thresholdand beam and walk-off beam turnedwalk-off the turned attention the attention to compensating to compensating the phase the velocityphase velocity dispersion dispersion by QPM. by QPM.QPM circumventsQPM circumvents the constraint the constraint of birefringence, of birefrin allowinggence, the allowing nonlinear the material nonlinear to be material engineered to be to produceengineered any to wavelength produce any within wavelength the transparency within the range transparency of the given range material, of the given and to material, take advantage and to ( ) (2) takeof the advantage largest element of the larg of theest nonlinearelement of susceptibility the nonlinear tensor susceptibilityχ 2 . Engineering tensor χ of. aEngineering material means of a materialto make itmeans to consist to make of a it periodical to consist structure of a periodic withal areas structure with alternatingwith areas oppositewith alternating crystallographic opposite (2) orientations in which the sign of the second order nonlinear susceptibility χ is opposite. The(2 first) crystallographic orientations in which the sign of the second order nonlinear susceptibility χ is practical realization of such a patterned structure was in ferroelectric periodically-poled LiNbO opposite. The first practical realization of such a patterned structure was in ferroelectric3 (PPLN). Strong intrinsic absorption in this material, however, limits the use of PPLN to wavelengths periodically-poled LiNbO3 (PPLN). Strong intrinsic absorption in this material, however, limits the shorter than 4 µm [40] (Figure9a). In non-ferroelectric materials, where it is not possible to pattern use of PPLN to wavelengths shorter than 4 µm [40] (Figure 9a). In non-ferroelectric materials, where the material through electrical field poling, QPM can be achieved by spatially inverting the nonlinear it is not possible to pattern the material through electrical field poling, QPM can be achieved by susceptibility during growth. Initially this was done through alternating the orientation of wafers in a spatially inverting the nonlinear susceptibility during growth. Initially this was done through stack [41]. However, high optical losses, observed at the wafer interfaces, and the small layer thickness alternating the orientation of wafers in a stack [41]. However, high optical losses, observed at the that is needed to satisfy the QPM conditions made these approaches unsuitable for practical devices. wafer interfaces, and the small layer thickness that is needed to satisfy the QPM conditions made The attempts to grow a thick layer on the edge of the stack, using the edge surface as a patterned these approaches unsuitable for practical devices. The attempts to grow a thick layer on the edge of template, solved the first issue, but not the second one. Advances in planar technology adopted from the stack, using the edge surface as a patterned template, solved the first issue, but not the second the microelectronic industry made possible the practical realization of thin micro-structured materials, one. Advances in planar technology adopted from the microelectronic industry made possible the called orientation-patterned (OP) templates, with the necessary domain reversal for QPM interaction. practical realization of thin micro-structured materials, called orientation-patterned (OP) templates, The last remaining step for making a QPM structure is to grow on the thin OP template a layer from the with the necessary domain reversal for QPM interaction. The last remaining step for making a QPM same material which is thick enough for a bulk frequency conversion device—maintaining the periodic structure is to grow on the thin OP template a layer from the same material which is thick enough for orientation of the initial template pattern is, of course, a strict requirement. GaAs with its broad IR a bulk frequency conversion device—maintaining the periodic orientation of the initial template transparency (Figure9a), high nonlinear optical susceptibility, and mature growth technology has been pattern is, of course, a strict requirement. GaAs with its broad IR transparency (Figure 9a), high highly successful as an OP nonlinear material [42,43]. However, OPGaAs is not without limitations, nonlinear optical susceptibility, and mature growth technology has been highly successful as an OP too, as the most unpleasant among all of them is, probably, its strong two-photon absorption (2PA) at nonlinear material [42,43]. However, OPGaAs is not without limitations, too, as the most unpleasant wavelengths below 1.7 µm [44]. This disadvantage cannot be neglected, because the range 1–1.7 µm is among all of them is, probably, its strong two-photon absorption (2PA) at wavelengths below 1.7 µm heavily populated with mature readily available commercial laser sources based on Nd, Yb, and Er [44]. This disadvantage cannot be neglected, because the range 1–1.7 µm is heavily populated with that could be otherwise used for pumping the GaAs QPM structure. In contrast, compared to GaAs, mature readily available commercial laser sources based on Nd, Yb, and Er that could be otherwise the same structured GaP has a negligible 2PA coefficient in the same spectral range (Figure9b). used for pumping the GaAs QPM structure. In contrast, compared to GaAs, the same structured GaP has a negligible 2PA coefficient in the same spectral range (Figure 9b).

Crystals 2017, 7, 178 12 of 38 Crystals 2017, 7, 178 12 of 38

Figure 9. ((aa)) IR transparency of some traditionaltraditional quasi-phase-matchingquasi-phase-matching (QPM) nonlinear optical materials [43]; [43]; (Reproduced (Reproduced from from [43] [43 with] with some some modifications modifications with with the permission the permission of SPIE); of SPIE);(b) A (comparisonb) A comparison of the two-photon of the two-photon absorption absorption (2PA) coefficient (2PA) coefficient β of GaAsβ ofand GaAs GaP andin the GaP near in IR the region near betweenIR region 1–1.7 between µm. 1–1.7 Initialµm. measurements Initial measurements of 3PA of 3PAin GaP in GaP (not (not shown shown here) here) indicate indicate almost undetectably low numbers [44]. [44]. (Rep (Reproducedroduced from [44] [44] with some modifications modifications with the permission of OSA).

GaP possesses also twice the thermal conductivity of of GaAs, GaAs, lower lower thermal thermal expansion expansion (Table (Table 11)) and aa broadbroad transparencytransparency range range that that starts starts conveniently conveniently in thein the visible visible region region [43] (Figure[43] (Figure9a)—the 9a)—the latter latterallows allows an easy an alignment easy alignment of the optics of the during optics the during related opticalthe related andnonlinear optical and optical nonlinear measurements. optical Allmeasurements. this is at the All cost this of ais slightly at the cost lower of a nonlinear slightly lower susceptibility. nonlinear susceptibility.

Table 1.1. AA comparisoncomparison between between some some optical optical and and thermal thermal properties properties of GaAs of GaAs and GaP.and 2PAGaP. = 2PAtwo-photon = two-photon absorption. absorption. Properties GaAs GaP Properties GaAs GaP Transmission window [µm] 1–16 0.55–12 Transmission window [µm] 1–16 0.55–12 −1 Nonlinear optical coefficient d14 [pm·V−1 ] 94 71 Nonlinear optical coefficient d14 [pm·V ] 94 71 IndexIndex of refractionrefractionn n 3.343.34 3.033.03 Thermal conductivityconductivity [W [W·m·m−1·−K1·K−1−]1] 55 55 110 110 − − ThermalThermal expansionexpansion coefficient coefficient [10 [106·K−6·K1]−1] 5.75.7 4.6 4.6 −1· −1 µ 2PA2PA coefficientcoefficient [cm/G [cm/G−1W·W−1] at 1.061.06 µmm 28 28 0.02 0.02

In addition, becausebecause GaPGaP hashas aa smallersmaller refractive refractive index index than than GaAs GaAs (Table (Table1), 1), the the QPM QPM period periodΛ calledΛ called also also “characteristic “characteristic length”, length”, will will be signiﬁcantly be significantly (about (about twice) twice) larger larger [45,46 [45,46]] at one at and one the and same the samenonlinear nonlinear interaction. interaction.

n 1 np nss ni == −− −− (6)(6) Λ λp λs λi p s i In this equation the indexes p, s and i denote the pump, signal, and idler waves, n is the refractive index,In andthis λequationis the wavelength. the indexes p , s and i denote the pump, signal, and idler waves, n is the refractiveAll these index, factors and madeλ is GaPthe wavelength. one of the most promising and studied QPM materials today. This led to theAll design these offactors the first made GaP GaP frequency one of the conversion most pr devicesomising basedand studied on stacked QPM GaP materials wafers today. [47] and This to ledthe to first the demonstration design of the first of QPM GaP parametricfrequency conver fluorescencesion devices in periodically based on stacked inverted GaP GaP wafers [48]. Recently[47] and toOPO the operation first demonstration was reported of in QPM OPGaP parametric grown on fluorescence native OPGaP in templates,periodically pumped inverted at GaP 2 µm [48]. [49] Recentlyand soon OPO after operation that at 1 µ mwas [50 reported] as the reportedin OPGaP OPGaP grown slope on native efficiency OPGaP was templates, 16% at 350 pumped mW output at 2 powerµm [49] for and the soon 2 µm after case, that and at much 1 µm less[50] for as the 1reportedµm case. OPGaP Thus the slope OPGaP efficiency performance was 16% still at 350 remains mW outputmodest power compared for the to 2 the µm 60% case, slope and much efficiency less routinelyfor the 1 µm obtained case. Thus in OPGaAs the OPGaP [51 ],performance although some still remainsnewer studies modest [52 compared] show further to the progress. 60% slope Obviously, efficiency improvements routinely obta inined the OPGaP in OPGaAs material [51], quality although are some newer studies [52] show further progress. Obviously, improvements in the OPGaP material

Crystals 20172017,, 77,, 178178 13 of 38

quality are necessary to reap the benefits of the low 2PA. It would be also a good idea to determine necessary to reap the beneﬁts of the low 2PA.(2) It would be also a good idea to determine again the GaP again the GaP nonlinear susceptibility χ , as the data quoted in the literature is relatively old [53]. nonlinear susceptibility χ(2), as the data quoted in the literature is relatively old [53]. Although remaining attractive and worth the effort, OPGaP started to show that it would not be Although remaining attractive and worth the effort, OPGaP started to show that it would not be an-easy-guy to deal with on the streets of the tough scientific neighborhood. an-easy-guy to deal with on the streets of the tough scientiﬁc neighborhood. Finally, while this work is focused mostly on the material aspects of quasi-phase-matching, the Finally, while this work is focused mostly on the material aspects of quasi-phase-matching, the readers who are interested in the device aspects of this research are kindly referred to the excellent, readers who are interested in the device aspects of this research are kindly referred to the excellent, comprehensive review on this matter by Hu et al. [54]. comprehensive review on this matter by Hu et al. [54].

6.2. First Disappointments with OPGaP

6.2.1. Additional Absorption Band In contrast to GaAs (Figure 1010a,a, the green line), GaP has an additional absorption band between 2–4 µµmm (Figure(Figure 1010a,a, thethe redred line).line). This absorption band cannot be neglected because it covers almost entirely one of the two atmospheric transparency windowswindows inin the IR region. Knowing the reasons for that,that, namely, thethe presencepresence ofof n-typen-type freefree carrierscarriers [[55,56],55,56], waswas notnot helpfulhelpful when buying GaP wafers, because all offered onon thethe marketmarket “undoped”“undoped” (rather(rather unintentionallyunintentionally doped) material which is only n-type, but it did help toto reducereduce thisthis absorptionabsorption [[57,58]57,58] laterlater duringduring thethe subsequentsubsequent thickthick growthgrowth (Figure(Figure 1010a,a,the the blue blue line). line). This This was was done done by by purposely purposely restricting restricting the incorporationthe incorporation of Si of coming Si coming from thefrom hot the reactor hot reactor quartz quartz parts. Moreover,parts. Moreover, by this theby this transparency the transparency of the grown of the GaP grown in the GaP entire in the IR regionentire wasIR region improved was makingimproved it closermaking to theit closer IR transparency to the IR transparency of the semi-insulating of the semi-insulating (SI) GaP (Figure (SI) 10GaPb), where(Figure the 10b), “bleaching” where the effect “bleaching” is achieved effect by is p-type achieved doping by p-type [59]. doping [59].

Figure 10.10. ((aa)) TheThe additional additional absorption absorption band band between between 2–4 2–4µm µm in GaPin GaP substrates substrates (red (red line) line) does does not exist not inexist GaAs in (greenGaAs line) (green and isline) reduced and in theis reduce Halided Vapor in the Phase Halide Epitaxy Vapor (HVPE) Phase heteroepitaxially Epitaxy (HVPE) grown GaPheteroepitaxially deposited on grown GaAs; GaP (b) the deposited HVPE grownon GaAs; material (b) the is HVPE as transparent grown material in the IR is as as the transparent commercial in p-dopedthe IR as SIthe GaP commercial [59]. (Reproduced p-doped SI from GaP [59 [59].] with (Reproduced the permission from of[59] OSA). with the permission of OSA).

6.2.2. Expensive but Low Quality Commercial GaPGaP GaP is not that popular as GaAs which explains why its fabrication process is not that mature. AsAs aa resultresult thethe GaPGaP marketmarket isis muchmuch narrower,narrower, whichwhichreﬂects reflects on on price price and and quality quality (Table (Table2). 2).

TableTable 2. AA comparison comparison of of price and quality (EPD/cm 2)2 )of of commercial commercial 2-inch 2-inch GaP GaP and and GaAs GaAs wafers. wafers.

Characteristics/WaferCharacteristics/Wafer Material Material GaP GaAs GaAs Price [USD] 585–685 87–90 Price [USD] 585–685 87–90 2 1 EPDEPD [etch [etch pits/cm pits/cm2] ] 80,000–100,00080,000–100,000 1500–50001500–5000 1 1 The1 dataThe datarepresent represent the the ranges ranges for for price price and and EPDEPD takentaken from from several several recent recent vendors’ vendors’ quotes. quotes. While the higher price of the substrates, as a major product component of the final product, will be a real problem later, the low surface quality is a serious obstacle right away. With such surface morphology one should expect nothing less than more intensive 3D growth in comparison to the

Crystals 2017, 7, 178 14 of 38

While the higher price of the substrates, as a major product component of the ﬁnal product, will be a real problem later, the low surface quality is a serious obstacle right away. With such surface morphology one should expect nothing less than more intensive 3D growth in comparison to the case of GaAs homoepitaxy. This is because the etch pits, being associated with ends of screw dislocations,Crystals provide 2017, 7, 178 plenty of sites for the atoms approaching the growing surface14 to of adhere38 and thus promote spiral growth and formation of 3D islands that due to the Ehrlich–Schwoebel effect case of GaAs homoepitaxy. This is because the etch pits, being associated with ends of screw will convertdislocations, in later growthprovide plenty stages of intosites largerfor the atoms 3D formations—hillocks. approaching the growing Thissurface means to adhere that and growth that starts on athus substrate promote surface spiral growth with higherand formation EPD willof 3D result islands in that a rougherdue to the surface Ehrlich–Schwoebel morphology effect (Figure 11). This effectwill is lessconvert pronounced in later growth if stages the bare into larger (plain) 3D substratesformations—hillocks. are with This some means miscut. that growth Then that the atomic terraces, formedstarts on duringa substrate the surface miscut, with providehigher EPD also will plenty result in of a rougher atomic surface sites formorphology the approaching (Figure atoms 11). This effect is less pronounced if the bare (plain) substrates are with some miscut. Then the and thus competeatomic terraces, with formed the etch during pits. the However, miscut, provide lower also quality plenty growthof atomic onsites a for bare the substrateapproaching also means a poorer surfaceatoms and morphology thus compete whenwith the the etch growth pits. However, is performed lower quality on growth OP templates on a bare substrate fabricated alsofrom such substrates.means Moreover, a poorer surface because morphology of this when reason the wegrowth should is performed exclude on OP one templates of the fabricated two known from template preparationsuch techniques substrates. Moreover, [60,61], because the wafer of this bonding reason we techniques, should exclude where one of the only two “on-axis” known template wafers can be preparation techniques [60,61], the wafer bonding techniques, where only “on-axis” wafers can be used for bonding,used for bonding, so a miscut so a miscut is not is not going going to to help. help.

Figure 11.FigureNomarski 11. Nomarski microscope microscope top top layer layer surfaces surfaces of of GaP GaP grown grown on plain on plain“on-axis” “on-axis” (100) GaP(100) (a) GaP (a) and (100) GaAsand (100) (b GaAs) substrates. (b) substrates. The The samples samples are are quartersquarters of of 2-inch 2-inch wafers. wafers. The images The imagesare taken are using taken using stitching software at 2.5× magnification [59]. (Reproduced from [59] with the permission of OSA). stitching software at 2.5× magniﬁcation [59]. (Reproduced from [59] with the permission of OSA). Another reason for excluding wafer bonding is the poor parallelism of the GaP Anotherwafers—bonding reason for excludingsuch wafers waferleaves voids bonding between is thethem poor (Figure parallelism 12a), someth ofing the that GaP usually wafers—bonding does not happen when bonding two GaAs wafers (Figure 12b). To avoid this, the GaP wafer parallelism such wafers leaves voids between them (Figure 12a), something that usually does not happen when should be improved before bonding by the so-called “fly-cut” technique [62]. This additional bonding twoprocedure, GaAs however, wafers (Figure not only 12 makesb). To this avoid template this, preparation the GaP wafermore expensive parallelism and complex, should but be improved before bondingalso leaves by several the so-called micron deep “ﬂy-cut” parallel grooves technique on the [ 62wafers’]. This surface additional that must be procedure, also polished however, off not only makesprior this to templatebonding. There preparation is another reason more to expensive give up the and wafer complex, bonding technique—the but also leaves fact that several to micron date a so-called “etch-stop” material for the case of GaP has not yet been discovered, while at least deep parallel grooves on the wafers’ surface that must be also polished off prior to bonding. There is two “etch-stoppers”, AlGaAs and InGaAs, can be used in the GaAs wafer bonding process. The another reasonmajor role to give of the up etch-stop the wafer is to bondingsecure the technique—thethickness of the so-called fact that invert toed date layer. a so-calledNo etch-stop “etch-stop” material forlayer the means case uncertainty of GaP has in notthe inverted yet been layer discovered, thickness and while in that at whether least two the “etch-stoppers”,subsequent HVPE AlGaAs and InGaAs,growth can has be in usedany way in theaccess GaAs to both wafer crystallographic bonding orientations, process. Theor only major to one role of them. of the More etch-stop is details can be found in [63]. to secure the thickness of the so-called inverted layer. No etch-stop layer means uncertainty in the Expensive, low quality GaP, an additional absorption band in one of the most important inverted layerfrequency thickness ranges, andabsence in that of any whether etch stop the materi subsequental means, HVPEas a result, growth giving has up inthe any cheaper way access to both crystallographicapproach for preparation orientations, of OP or templates. only to one Finally, of them. low template More detailsquality canand, bepresumably, found in poor [63 ]. Expensive,quality lowof the quality following GaP, thick an growth additional on them absorption are also an band issue. in Can one all of this the be most traded important off for the frequency lower 2PA, higher thermal conductivity, and transparency range that starts conveniently in the ranges, absencevisible? It of was, any maybe, etch time stop to materialstart looking means, for alternative as a result,substrate giving materials, up asthe well cheaperas growth approachand for preparationtemplate of OP preparation templates. approaches. Finally, low template quality and, presumably, poor quality of the following thick growth on them are also an issue. Can all this be traded off for the lower 2PA, higher thermal conductivity, and transparency range that starts conveniently in the visible? It was, maybe, time to start looking for alternative substrate materials, as well as growth and template preparation approaches. Crystals 2017, 7, 178 15 of 38 Crystals 2017, 7, 178 15 of 38

Crystals 2017, 7, 178 15 of 38

Figure 12. (a) Due to poor parallelism 2 two-inch bonded “on-axis” (100) GaP wafers indicate a large Figure 12. (a) Due to poor parallelism 2 two-inch bonded “on-axis” (100) GaP wafers indicate a large useless area of voids between the wafers; (b) Such a problem does not exist when 2 GaAsGaAs wafers are Figure 12. (a) Due to poor parallelism 2 two-inch bonded “on-axis” (100) GaP wafers indicate a large bonded [59]. (Reproduced from [59] with the permission of OSA). bondeduseless [59]. area (Reproduced of voids between from the[59] wafers; with the (b) permissionSuch a problem of OSA). does not exist when 2 GaAs wafers are bonded [59]. (Reproduced from [59] with the permission of OSA). 6.3. Alternative Substrate Materials 6.3. Alternative Substrate Materials Lattice mismatch between the substrate and the growinggrowing material—thematerial—the smaller the better—isbetter—is the mostmost Lattice importantimportant mismatch factor between during the heteroepitaxy. substrate and This the growing explains material—the the choice of smaller thethe ﬁrstfirst the heteroepitaxialheteroepitaxial better—is attemptthe most [[64]64] ofimportant growthgrowth factor onon OPOP during templates,templates, heteroepitaxy. particularlyparticularly This explains ofof OPZnSeOPZnSe the choice onon OPGaAsOPGaAs of the first template—ZnSetemplate—ZnSe heteroepitaxial and attempt [64] of growth on OP templates, particularly of OPZnSe on OPGaAs template—ZnSe and GaAs stay almost on a verticalvertical line,line, asas shownshown onon FigureFigure 1313b.b. Following thisthis logic the close match GaAs stay almost on a vertical line, as shown on Figure 13b. Following this logic the close match between SiSi and and GaP GaP could could make make Si a Si perfect a perfect substrate subs fortrate heteroepitaxy for heteroepitaxy of GaP. of Surprisingly, GaP. Surprisingly, however, between Si and GaP could make Si a perfect substrate for heteroepitaxy of GaP. Surprisingly, our attempts to grow GaP on Si resulted only in a great number of small GaP crystallites, distributed however,however, our ourattempts attempts to togrow grow GaP GaP on on Si Si resulted resulted onlyonly in in a agreat great number number of smallof small GaP GaP crystallites, crystallites, randomlydistributeddistributed on randomly the randomly Si-substrate on onthe the Si-substrate surface Si-substrate (Figure surface surface 13a). (Figure 13a). 13a).

FigureFigure 13. ( a13.) HVPE (a) HVPE growth growth of GaPof GaP directly directly on on Si; Si; (b ()b Bandgap) Bandgap energy energy vsvs latticelattice constant constant for for some some of of the mostthe popular most semiconductorpopular semiconductor materials materials (Reproduced (Reproduced from [65 from] with [65] some with modifications); some modifications); (c) Physical Figure 13. (a) HVPE growth of GaP directly on Si; (b) Bandgap energy vs lattice constant for some of vapor(c) transport Physical vapor growth transport of OPZnSe growth on of aOPZnSe OPGaAs on a template OPGaAs template [64]. (Reproduced [64]. (Reproduced from from [64] with[64] the the most popular semiconductor materials (Reproduced from [65] with some modifications); permissionwith the of permissionJ. Cryst. Growth of J. Cryst.). Growth). (c) Physical vapor transport growth of OPZnSe on a OPGaAs template [64]. (Reproduced from [64] with Similarly,the permission the Physical of J. Cryst. Vapor Growth Transport). (PVT) growth of OPZnSe on OPGaAs templates led to Similarly,poor domain the propagation Physical Vapor (Figure Transport 13c) and limited (PVT) optical growth results. of OPZnSe on OPGaAs templates led to poorSimilarly, domainThese propagation ratherthe Physical discouraging (Figure Vapor than 13 Transportc) optimistic and limited (PVT) resu opticallts gr showedowth results. of that, OPZnSe probably, on OPGaAsheteroepitaxy templates was not led to poorThese goingdomain to rather do propagation the discouraging job even (Figurein such than favorable13c) optimistic and cases,limited results which optical showednaturally results. that,influenced probably, the research, heteroepitaxy directing was it not goingThesetowards to do ratherthe development job discouraging even inof suchtechniques than favorable optimistic for the cases, preparation resu whichlts showed naturally of native that, influencedOPGaP probably, template. theheteroepitaxy research, At this point directing was not itgoing towardsgrowth to do development theof GaP job on even GaAs in of atsuch techniques much favorable larger for lattice cases, the mismatch preparation which naturally(Figure of native 13b) influenced sounded OPGaP like the template. wasting research, money At directing this in point it a hopeless adventure. That is why we forgot our first timid attempt (for more details see Section growthtowards of development GaP on GaAs of at techniques much larger for lattice the mismatchpreparation (Figure of native 13b) soundedOPGaP template. like wasting Atmoney this point in a 6.7.1) to grow OPGaP on the only available OPGaAs templates at that time and proceeded as the rest hopelessgrowth of adventure. GaP on GaAs That at is much why larger we forgot lattice our mismatch first timid (Figure attempt 13b) (for sounded more details like wasting see Section money 6.7.1 in)

a hopeless adventure. That is why we forgot our first timid attempt (for more details see Section 6.7.1) to grow OPGaP on the only available OPGaAs templates at that time and proceeded as the rest

Crystals 2017, 7, 178 16 of 38

toCrystals grow OPGaP2017, 7, 178 on the only available OPGaAs templates at that time and proceeded as the16 restof 38 of the world in a widely accepted direction. However, meeting all the aforementioned difficulties with GaPof severalthe world years in a later widely we resumedaccepted ourdirection. attempts Howe onver, GaP/GaAs meeting heteroepitaxy—GaPall the aforementioned and difficulties GaAs were, andwith still GaP are several all the same,years later the mostwe resumed promising our candidatesattempts on for GaP/GaAs frequency heteroepitaxy—GaP conversion in the and IR and GaAs THz region.were, Plus,and still the are OPGaAs all the templatesame, the preparation, most promisin ing contrast candidates to the for preparationfrequency conversion of OPGaP in templates the IR wasand already THz region. for many Plus, years the atOPGaAs a mature template stage. preparation, in contrast to the preparation of OPGaP templates was already for many years at a mature stage. Surprisingly, from the first experiments the GaP/GaAs heteroepitaxy resulted in the same smooth Surprisingly, from the first experiments the GaP/GaAs heteroepitaxy resulted in the same surface morphology in comparison to the surface morphology after GaP/GaP homoepitaxy, with smooth surface morphology in comparison to the surface morphology after GaP/GaP homoepitaxy, the same typical growth features for a 2D and an average surface roughness (rms) within 1 nm in with the same typical growth features for a 2D and an average surface roughness (rms) within 1 nm 1 µm × 1 µm square Atomic-force microscopy (AFM) images (Figure 14a). The crystalline quality, in 1 µm × 1 µm square Atomic-force microscopy (AFM) images (Figure 14a). The crystalline quality, expressedexpressed by by the the Full Full Width Width of of Half Half MaximumMaximum (FWHM) of of an an omega-scan omega-scan reflection reflection signal signal was was also also comparablecomparable for for both both GaP/GaP GaP/GaP andand GaP/GaAsGaP/GaAs cases cases (Figure (Figure 14b), 14b), although although the the initial initial crystalline crystalline qualityquality of of the the GaAs GaAs substrate substrate (before (before growth) growth) waswas higherhigher (Figure 14b,c).14b,c).

Figure 14. (a) Atomic-force microscopy (AFM) images (1 µm × 1 µm square) of HVPE growth of GaP Figure 14. (a) Atomic-force microscopy (AFM) images (1 µm × 1 µm square) of HVPE growth of GaP on GaAs show smooth surface morphology; (b) Full Width of Half Maximum (FWHM) omega-scan on GaAs show smooth surface morphology; (b) Full Width of Half Maximum (FWHM) omega-scan of GaP and GaAs substrate surfaces before growth and the related results after HVPE homo- and of GaP and GaAs substrate surfaces before growth and the related results after HVPE homo- and heteroepitaxy [59]. (Reproduced from [59] with the permission of OSA); (c) Reflection curves of heteroepitaxy [59]. (Reproduced from [59] with the permission of OSA); (c) Reflection curves of omega-scans of a GaP and a GaAs substrate before the HVPE growth. omega-scans of a GaP and a GaAs substrate before the HVPE growth. The heteroepitaxial experiments included HVPE growths of GaP on a GaAs substrate (FigureThe heteroepitaxial 15a) and vice versa experiments of GaAs included on a GaP HVPE substrate. growths The of growth GaP on of a GaP GaAs on substrate GaAs resulted (Figure in 15 a) andbetter vice surface versa of morphology GaAs on a GaPand higher substrate. crystalline The growth quality. of This GaP means on GaAs that resulted the sign in of better the misfit surface morphology(mismatch) anddoes higher matter, crystalline as at a negative quality. misfit This (growth means that of a the material sign of with the misfita smaller (mismatch) lattice on does a matter,substrate as at with a negative a larger misfit lattice) (growth seems to of be a materialmore favorable. with a smallerIn such a lattice case the on growing a substrate layer with must a larger be lattice)under seems a tensile to bestrain more in favorable.contrast to Inthe such opposite a case case the (positive growing misfit) layer mustwhen bethe under strain ais tensile compressed. strain in contrastLooking to theat the opposite graph on case Figure (positive 13b, misfit)this means when that the when strain we is compressed.make a choice Looking for heteroepitaxy at the graph the on Figuregrowing 13b, material this means must that be whenat the weleft makeof the asubstrat choicee. for This heteroepitaxy is exactly the the case growing of GaP materialgrown on must GaAs. be at The experiments were conducted with and without the deposition of an intermediate buffer the left of the substrate. This is exactly the case of GaP grown on GaAs. layer between the substrate and the growing layer. The intermediate buffer layer aimed to The experiments were conducted with and without the deposition of an intermediate buffer layer accommodate the growing layer to the substrate by, possibly, a gradual change of the lattice between the substrate and the growing layer. The intermediate buffer layer aimed to accommodate constants. In the experiment this layer was formed by starting the growth with an entirely arsine the growing layer to the substrate by, possibly, a gradual change of the lattice constants. In the flow, which was gradually changed in the mixture AsH3/PH3 to an entirely phosphine flow. experiment this layer was formed by starting the growth with an entirely arsine flow, which was However, as the elemental profile analysis across the interface indicated (Figure 15b), no matter graduallywhether changed we are going in the to mixture form or AsH not3 /PHan intermedia3 to an entirelyte layer, phosphine such a layer flow. always However, exists, as thedue elemental to the profilemutual analysis diffusion across at the the interface interface of indicated As and, re (Figurespectively, 15b), P noatoms matter towards whether both wegrowing are going layer to and form orsubstrate. not an intermediate Moreover, when layer, the such ratio a layer of the always AsH3 exists,/PH3 stayed due to constant, the mutual it was diffusion still necessary at the interface some oftime As and, for the respectively, As and P atoms P atoms to replace towards each both other growing before the layer same and ratio substrate. could be Moreover,established whenwithin the the crystal cell. In this process, as the study showed, the structure stayed crystalline all the way along

Crystals 2017, 7, 178 17 of 38

ratio of the AsH3/PH3 stayed constant, it was still necessary some time for the As and P atoms to replace each other before the same ratio could be established within the crystal cell. In this process, as Crystals 2017, 7, 178 17 of 38 the studyCrystals showed, 2017, 7, 178the structure stayed crystalline all the way along these replacements. Thus,17 of 38 with some practice we were able to start, for example, with an HVPE growth of GaAs on a GaAs substrate these replacements. Thus, with some practice we were able to start, for example, with an HVPE these replacements. Thus, with some practice we were able to start, for example, with an HVPE and aftergrowth a whileof GaAs to on continue a GaAs substrate in a one-step-growth and after a while process to continue with in HVPE a one-step-growth growth of GaP process (Figure with 16a). growth of GaAs on a GaAs substrate and after a while to continue in a one-step-growth process with By extendingHVPE growth these of ideas GaP (Figure we were 16a). also By able extending to grow thes ate certainideas we AsH were3/PH also3 ableratios to grow good at quality certain thick HVPE growth of GaP (Figure 16a). By extending these ideas we were also able to grow at certain GaAsAsHxP1−3/PHx ternaries3 ratios good (Figure quality 16b) thick expecting GaAsxP1 to−x ternaries combine (Figure the best 16b) properties expecting to of combine these two thematerials best AsH3/PH3 ratios good quality thick GaAsxP1−x ternaries (Figure 16b) expecting to combine the best relatedproperties to pursuit of application,these two materials namely, therelated larger to nonlinearpursuit application, susceptibility namely, of GaAs the withlarger the nonlinear smaller 2PA properties of these two materials related to pursuit application, namely, the larger nonlinear susceptibility of GaAs with the smaller 2PA of GaP. of GaP.susceptibility of GaAs with the smaller 2PA of GaP.

FigureFigure 15. ( a15.) Cross (a) Cross section section of of HVPE HVPE heteroepitaxy heteroepitaxy of GaP GaP on on GaAs GaAs substrates; substrates; (b) (Elementalb) Elemental profile proﬁle Figure 15. (a) Cross section of HVPE heteroepitaxy of GaP on GaAs substrates; (b) Elemental profile analysisanalysis across across the the interface interface between between the the GaAs GaAs substratesubstrate and and the the grown grown GaP GaP layer layer [59]. [59 (Reproduced]. (Reproduced analysis across the interface between the GaAs substrate and the grown GaP layer [59]. (Reproduced from [59] with the permission of OSA). fromfrom [59] with[59] with the permissionthe permission of of OSA). OSA).

Figure 16. (a) A cross section image of a combined HVPE growth, which starts first with a FigureFigure 16. ( a16.) A ( crossa) A sectioncross section image image of a combined of a combined HVPE growth,HVPE growth, which startswhich ﬁrst starts with first a GaAs/GaAswith a GaAs/GaAs homoepitaxy and after that continues with GaP/GaAs heteroepitaxy; (b) HVPE growth homoepitaxyGaAs/GaAs and homoepitaxy after that continues and after withthat cont GaP/GaAsinues with heteroepitaxy; GaP/GaAs heteroepitaxy; (b) HVPE growth (b) HVPE of GaAsgrowth03 P0.66 of GaAs03P0.66 on a GaP template. on a GaPof GaAs template.03P0.66 on a GaP template. 6.4. The HVPE Growth Process 6.4. The6.4. HVPE The HVPE Growth Growth Process Process The traditional Hydride Vapor Phase Epitaxial (HVPE) process (called “Hyrdide” in the case of The traditional Hydride Vapor Phase Epitaxial (HVPE) process (called “Hyrdide” in the case of TheGaAs traditional or GaP growth Hydride but “Halide” Vapor when Phase we Epitaxial grow III-Nitrides) (HVPE) processcan be performed (called “Hyrdide” in a horizontal in thehot case GaAs or GaP growth but “Halide” when we grow III-Nitrides) can be performed in a horizontal hot wall quartz reactor (Figure 17a). Although that it is a relatively old growth technique, the HVPE, in of GaAswall or quartz GaP reactor growth (Figure but “Halide” 17a). Although when that we growit is a relatively III-Nitrides) old growth can be technique, performed the in HVPE, a horizontal in contrast to its modern alternatives well-known in the semiconductor industry, the Metal Organic hot wallcontrast quartz to its reactor modern (Figure alternatives 17a). Although well-known that in itthe is semiconductor a relatively old industry, growth the technique, Metal Organic the HVPE, Chemical Vapor Deposition (MOCVD) and the Molecular Beam Epitaxy (MBE) is the only method in contrastChemical to itsVapor modern Deposition alternatives (MOCVD) well-known and the Molecular in the semiconductor Beam Epitaxy (MBE) industry, is the the only Metal method Organic for a relatively fast (hundreds of µm/h), and thick (up to millimeters) growth of binary Chemicalfor a Vapor relatively Deposition fast (hundreds (MOCVD) of andµm/h), the Molecularand thick Beam(up to Epitaxy millimeters) (MBE) growth is the only of methodbinary for semiconductor materials such as GaN, GaP, GaAs, etc., as well as their ternaries and quaternaries. In semiconductor materials suchµ as GaN, GaP, GaAs, etc., as well as their ternaries and quaternaries. In a relativelythe particular fast (hundreds case of HVPE of growthm/h), of and GaP thick a quartz (up boat to millimeters) with molten growthgallium is of placed binary near semiconductor the inlet the particular case of HVPE growth of GaP a quartz boat with molten gallium is placed near the inlet materialsof a horizontal such as GaN, quartz GaP, tube GaAs, (reactor etc., tube), as well which as is their placed ternaries within anda 3- quaternaries.or 4-zone resistive In the furnace. particular of a horizontal quartz tube (reactor tube), which is placed within a 3- or 4-zone resistive furnace. case ofPrecisely, HVPE growththe Ga-boat, of GaP is inside a quartz of a boatsmaller with quartz molten tube, gallium called a is nozzle, placed which near theis positioned inlet of a horizontalin the Precisely, the Ga-boat, is inside of a smaller quartz tube, called a nozzle, which is positioned in the quartzmiddle tube (reactoralong the tube), axis of which the reactor is placed tube. within A mixture a 3- or of 4-zone HCl and resistive H2 overflows furnace. from Precisely, the Ga theboat Ga-boat, as middle along the axis of the reactor tube. A mixture of HCl and H2 overflows from the Ga boat as is insideHCl ofpicks a smaller up some quartz of the tube, molten called Ga, forming a nozzle, with which different is positioned probabilities in theGaCl, middle GaCl2 alongor GaCl the3, axis HCl picks up some of the molten Ga, forming with different probabilities GaCl, GaCl2 or GaCl3,

Crystals 2017, 7, 178 18 of 38

of the reactor tube. A mixture of HCl and H2 overflows from the Ga boat as HCl picks up some of the molten Ga, forming with different probabilities GaCl, GaCl2 or GaCl3, releasing at the same time an additionalCrystals 2017 amount, 7, 178 of H2. The role of the H2 flow through the nozzle is not only to carry18 of 38 but also to dilute the flow of HCl to a desired extent. Once formed, gallium chloride is further delivered out releasing at the same time an additional amount of H2. The role of the H2 flow through the nozzle is of the nozzle.not only Anotherto carry but flow also to of dilute phosphine the flow PHof HCl3 (also to a desired carried extent. and dilutedOnce formed, by a gallium separate chloride H2 flow) is introducedis further separately delivered at out the of periphery the nozzle. ofAnother the reactor flow of tube.phosphine The PH33 (alsoflow carried meets and for diluted the first by a time the galliumseparate chloride H flow2 flow) at is the introduced end of the separately nozzle at tube. the periphery The two of gases the reactor make atube. mixture The PH with3 flow the meets intention to form GaP,for hopefully,the first time in the surface gallium reactions chloride thatflow take at the place end onof the the nozzle growing tube. substrate The two gases or template make a surface placed inmixture the way with of the the intention mixture to (Figure form GaP,17a). hopefull However,y, inearlier surface formation reactions that of GaP,take forplace example, on the on the growing substrate or template surface placed in the way of the mixture (Figure 17a). However, edge of the nozzle or on the inner reactor wall is also possible—we will call these reactions “parasitic” earlier formation of GaP, for example, on the edge of the nozzle or on the inner reactor wall is also (Figure possible—we17b). will call these reactions “parasitic” (Figure 17b).

Figure 17. (a) HVPE growth of GaP in a low-pressure horizontal quartz reactor; (b) parasitic Figure 17. (a) HVPE growth of GaP in a low-pressure horizontal quartz reactor; (b) parasitic nucleation nucleation on the quartz tube, luckily in the area that is behind the substrate [63]. (Reproduced from on the quartz[63] with tube, the permission luckily in of the SPIE). area that is behind the substrate [63]. (Reproduced from [63] with the permission of SPIE). The forming of GaP relies on reactions of the type: + = + The forming of GaP relies on reactionsGaCl3 ofPH the3 type:GaP 3HCl (7)

To take place this reaction needs some atomic Ga and P but they both are combined in GaCl + PH = GaP + 3HCl (7) molecules. How then does the chemistry3 work? 3When gallium chloride reaches the substrate surface, if there is some atomic hydrogen H available, this atomic hydrogen will readily combine with the ToCl-atom take place from this the reaction GaCl molecule needs (forming some atomic HCl) and, Ga andthus, P will but liberate they both the Ga are atom. combined This process in molecules. How thencalled does “dechlorination” the chemistry of the work? substrate When surface gallium is an chlorideimportant reachesstep in forming the substrate GaP. The surface,problem, if there is somehowever, atomic hydrogenis, again, that H the available, whole amount this atomicof carrier hydrogen H2 gas is in will the form readily of molecular combine hydrogen with the H2 Cl-atom. from theHydrogen GaCl molecule is as much (forming as you HCl)want and,in the thus, stream will but, liberate regretfully, the Gathe atom.growth This temperature process called “dechlorination”(720–740 °C) of is the not substrate high enough surface to disintegrate is an important the hydrogen step in molecule forming into GaP. 2 Thehydrogen problem, atoms. however, Fortunately, the same temperature is high enough to disintegrate the arriving phosphine molecules is, again, that the whole amount of carrier H gas is in the form of molecular hydrogen H . Hydrogen PH3. As a result at each single act of disintegration2 we will have “liberated” one phosphorus 2atom ◦ is as muchand three as you hydrogen want atoms. in the What stream is better but, than regretfully, that? The thereleased growth atomic temperature hydrogen will (720–740 dechlorinateC) is not high enoughthe surface to disintegrate liberating some the Ga hydrogen atoms. At molecule the same into time 2 the hydrogen atomic phosphorus atoms. Fortunately, from the PH the3 same temperaturedisintegration is high enoughwill be ready to disintegrate to join the free the Ga arriving atoms forming phosphine in this moleculesway GaP. Of PH course,3. As these a result two at each single actshould of disintegration be “in-a-hurry” we because will havethere “liberated”is always a risk one some phosphorus of the free atomhydrogen and atoms three will hydrogen reform atoms. hydrogen molecules again, or in the same way, some of the phosphorus atoms will form P2 What is better than that? The released atomic hydrogen will dechlorinate the surface liberating some molecules, or even the more stable complex P4. This scenario (Figure 18) aims to show that the Ga atoms. At the same time the atomic phosphorus from the PH3 disintegration will be ready to join the free Ga atoms forming in this way GaP. Of course, these two should be “in-a-hurry” because there is always a risk some of the free hydrogen atoms will reform hydrogen molecules again, or in the Crystals 2017, 7, 178 19 of 38

same way, some of the phosphorus atoms will form P2 molecules, or even the more stable complex P4. Crystals 2017, 7, 178 19 of 38 This scenario (Figure 18) aims to show that the process of formation of GaP on the substrate surface is not easyprocess and can of beformation accompanied of GaP byon athe number substrate of unpleasantsurface is not events easy and and can parasitic be accompanied reactions by that a put at risk the wholenumber mission.of unpleasant events and parasitic reactions that put at risk the whole mission.

Figure 18. Illustration of the most probable growth chemistry of GaP growing by HVPE in a Figure 18. Illustration of the most probable growth chemistry of GaP growing by HVPE in a low-pressure horizontal quartz reactor. low-pressure horizontal quartz reactor. Things become even worse if the grower does not succeed in suppressing the parasitic Thingsnucleation become that even may worsehappen if upfront the grower at the doessubstrate not in succeed the mixing in suppressing area, or even thestill parasiticon the nozzle nucleation edge, where gallium chloride and phosphine for the first time “see” each other. Then this parasitic that may happen upfront at the substrate in the mixing area, or even still on the nozzle edge, where nucleation, stealing from the flows, will weaken the local supply of chemicals needed for the real galliumgrowth chloride on the and substrate phosphine surface, for and the produce first time in the “see” same eachor similar other. reaction Then (6) this additional parasitic HClnucleation, that stealingmay from start the etching flows, back will the weaken already the grown local surface supply and, of finally, chemicals will change needed over for time the the real V/III growth ratio on the substratewhich, surface, as it andwas producepointed out in by the many same authors or similar [63,66], reaction plays an (6) important additional role HClin these that processes. may start etching back the already grown surface and, finally, will change over time the V/III ratio which, as it was 6.5. Combining Two Growth Approaches pointed out by many authors [63,66], plays an important role in these processes. Combining two growth approaches could be as useful as combining two materials. For 6.5. Combiningexample, Two as a Growth close-to-equilibrium Approaches process HVPE provides about equal probabilities for formation and disintegration, i.e., returning to the gas phase, of a nucleus. Performed typically at low Combiningsupersaturation two growth levels approachesHVPE, in addition, could berelies as useful mostly as on combining surface defects, two materials. such as Forscrew example, as a close-to-equilibriumdislocations, atomic terraces, process etc., HVPE to initiate provides nucleation about processes. equal probabilitiesIt is just energetically for formation more and disintegration,favorable i.e., for returning the approaching to the gasatom phase, to adhere of a nucleus.at a surface Performed defect than typically directly aton lowa new supersaturation crystal levels HVPE,surface in (see addition, Section relies2). Thus mostly the better on surfacethe substr defects,ate surface such means as screw the lesser dislocations, the chances atomic for the terraces, HVPE to do the job. We realized that, namely, this was the real reason for the failure of our first etc., to initiate nucleation processes. It is just energetically more favorable for the approaching atom to attempts to grow GaP by HVPE directly on Si (Figures 13a and 19a)—the readily available, fairly adhere atcheap a surface silicon defectwafers thanpossessed directly extremely on a newhigh crystalquality surface(a typical (see EPD Section < 100/cm2).2—compare Thus the betterthis the substratenumber surface with means the EPD the lesserof GaAs the and chances GaP from for Table the HVPE 2). This to means do the that job. the We Si is realized just “too that, good” namely, for this was the realHVPE reason growth. for In the contrast, failure other of our growth first attemptstechniques, to such grow as GaPMOCVD by HVPE or MBE directly that provide on Si better (Figures 13a and 19a)—theconditions readily for 2D available, growth (like fairly high cheap supersaturatio silicon wafersn) and are possessed far from extremelyequilibrium, high would quality not have (a typical EPD < 100/cmsuch problems.2—compare Thus combining this number two with growth the processe EPD ofs, GaAs one close and to GaP and fromone far Table from2 ).equilibrium, This means that could be very useful. This was, actually, realized long ago and has become a daily routine for many the Si isyears just “toonow in good” the III-Nitride for HVPE growth growth. practice, In contrast, where, for other example, growth thick techniques,HVPE growth such of GaN as MOCVDis or MBEperformed that provide on a better thin MOCVD conditions or MBE for 2DAlN growth layer deposited (like high in supersaturation)advance on a sapphire and or are SiC far from equilibrium,substrate. would This not convinced have such us to problems.adopt this approach Thus combining and resume two our growthHVPE growth processes, of GaP one on Si, close this to and one far fromtime on equilibrium, a 1–2 µm thin could GaP bedeposited very useful. in advanc Thise by was, theactually, MOCVD realizedtechnique long on the ago Si-wafer. and has Thus become a daily routinehundreds for of many microns years thick now GaP in was the successful III-Nitridely grown growth by HVPE practice, on Si where, (Figure for19c). example, thick HVPE growth of GaN is performed on a thin MOCVD or MBE AlN layer deposited in advance on a sapphire or SiC substrate. This convinced us to adopt this approach and resume our HVPE growth of GaP on Si, this time on a 1–2 µm thin GaP deposited in advance by the MOCVD technique on the Si-wafer. Thus hundreds of microns thick GaP was successfully grown by HVPE on Si (Figure 19c). Crystals 2017, 7, 178 20 of 38 Crystals 2017, 7, 178 20 of 38 Crystals 2017, 7, 178 20 of 38

FigureFigure 19. 19.(a )(a HVPE) HVPE growth growth of of GaP GaP directlydirectly onon SiSi (duplicates Figure Figure 13a); 13a); (b (b) )Nomarski Nomarski top top surface surface imagesimagesFigure of 19.of a direct a( adirect) HVPE homoepitaxial homoepitaxial growth of GaP HVPE HVPE directly growthgrowth on ofSi GaP (duplicates on on a a GaP GaP Figure substrate substrate 13a); (upper(b (upper) Nomarski image) image) topand and surface HVPE HVPE GaPGaPimages grown grown of ona ondirect Metal Metal homoepitaxial Organic Organic Chemical Chemical HVPE Vapor Vaporgrowth DepositionDeposition of GaP on (MOCVD)a GaP substrate GaP/Si; GaP/Si; (upper (c ()c SEM) SEMimage) cross cross and section section HVPE of of HVPEHVPEGaP GaPgrown GaP grown grownon Metal on on MOCVD OrganicMOCVD Chemical GaP/Si. GaP/Si. Vapor Deposition (MOCVD) GaP/Si; (c) SEM cross section of HVPE GaP grown on MOCVD GaP/Si. ThisThis combined combined hetero-MOCVD/homo-HVPE hetero-MOCVD/homo-HVPE approach approach resulted resulted in anin extremely an extremely smooth smooth “like-Si” “like-Si”This GaP combined surface morphologyhetero-MOCVD/homo-HVPE (Figure 19b, lower approachimage) where resulted even thein GaP/GaPan extremely homoepitaxy smooth GaP surface morphology (Figure 19b, lower image) where even the GaP/GaP homoepitaxy “orange “orange“like-Si” peel”GaP surfacesurface morphologytexture typical (Figure for HVPE 19b, lois wernot presentedimage) where (Figure even 19b, the upper GaP/GaP image). homoepitaxy peel” surface texture typical for HVPE is not presented (Figure 19b, upper image). “orange peel” surface texture typical for HVPE is not presented (Figure 19b, upper image). 6.6.6.6. Purpose, Purpose, Preparation Preparation and and Growth Growth onon Half-PatternedHalf-Patterned Templates 6.6. Purpose, Preparation and Growth on Half-Patterned Templates In crystal growth, if something is a disadvantage at given circumstances, it also could be a great In crystal growth, if something is a disadvantage at given circumstances, it also could be a advantageIn crystal at others. growth, For if something example, theis a factdisadvantage that HVPE at isgiven a close-to-equilibrium circumstances, it also process could madebe a great our great advantage at others. For example, the fact that HVPE is a close-to-equilibrium process made attemptadvantage for directat others. HVPE For growth example, of GaP the on fact Si unsuccessful.that HVPE is However, a close-to-equilibrium namely, this gave process us the made chance our to our attempt for direct HVPE growth of GaP on Si unsuccessful. However, namely, this gave us growattempt on for the direct so-called HVPE half-patterned growth of GaP (HP) on templates. Si unsuccessful. This is However,a kind of a namely, selective this growth gave that us the was chance already to thediscussedgrow chance on the toin so-calledSection grow on3.5 half-patterned thein correlation so-called (HP) to half-patterned ELO templates. and PE ThisNDEO (HP) is a kindepitaxy templates. of aof selective III-Nitrides This growth is a[67] kind that (Figures was of a already selective7 and growth20a),discussed although that in was Section this already selective 3.5 in discussed correlation growth inhasto SectionELO been and st 3.5udied PE inNDEO for correlation many epitaxy other of to III-Nitrides ELOmaterials and (for PENDEO[67] example, (Figures epitaxy 7GaAs and of III-Nitrides(Figure20a), although 20b)), [67] as (Figuresthis well. selective 7 and growth 20a), although has been thisstudied selective for many growth other has materials been studied (for example, for many GaAs other materials(Figure (for20b)), example, as well. GaAs (Figure 20b)), as well.

Figure 20. (a) Cross sectional images of GaN selective growth performed at different crystallographic FiguredirectionsFigure 20. 20.(a ()anda Cross) Cross growth sectional sectional conditions images images [67] ofof GaNGaN(Repro selectiveduced fromgrowth growth [67] performed performed with some at at modifications);different different crystallographic crystallographic (b) Study directionsselectivedirections andgrowth and growth growth of GaAs conditions conditions to determine [[67]67] (Reproduced(Repro the influenceduced fromof growth [67] [67] with with temperature some some modifications); modiﬁcations); and V/III ratio (b) ( bonStudy) Studythe selectivemesas’selective growthshape growth [68]. of of GaAs(Reproduced GaAs to to determine determine from [68] thethe with inﬂuenceinfluence the permission of growth of J.temperature temperature Cryst. Growth and and). V/III V/III ratio ratio on onthe the mesas’mesas’ shape shape [68 [68].]. (Reproduced (Reproduced from from [[68]68] withwith thethe permission of of J.J. Cryst. Cryst. Growth Growth). ). To grow selectively means to grow on a patterned template, where some areas are open for growth,ToTo grow growwhereas selectively selectively others are means means prohibited toto growgrow from onon growth. aa patternedpatterned To prohibit template, template, the wheregrowth where some is some relatively areas areas are easy are open openby thefor for growth,growth, whereas whereas others others areare prohibited from from growth. growth. To Toprohibit prohibit the thegrowth growth is relatively is relatively easy by easy the by

Crystals 2017, 7, 178 21 of 38 Crystals 2017, 7, 178 21 of 38 thedeposition deposition of of a a thin thin “unfriendly” toto thethe growthgrowth material,material, such such as as SiO SiO2 or2 or Si 3NSi43N.4 Then. Then the the close-to-equilibriumclose-to-equilibrium nature nature of of HVPE HVPE will will do do the the rest rest of of the the job—even job—even if aifnucleus a nucleus has has the the chance chance to to crystallizecrystallize on theon prohibitedthe prohibited area, it willarea, be it easily will disintegratedbe easily disintegrated due to the relatively due to low the supersaturation relatively low levelssupersaturation maintained during levels themaintained HVPE process. during By the studying HVPE process. the selective By studying growth by the the selective deposition growth of the by patternthe deposition in a different of the crystallographic pattern in a different orientation crysta onellographic can figure orientation out important one can growth figure characteristic out important suchgrowth as mesas’ characteristic shape, growth such as rate mesas’ in different shape, growth crystallographic rate in different directions, crystallographic etc. and, thus, directions, have an etc. accurateand, thus, control have on an the accurate growth morphology.control on the Usually, growth no morphology. matter which Usually, is the studied no matter material, which there is the is alwaysstudied amaterial, case when there the is mesas’ always shape a case is when rectangular the mesas’ as often shape this is isrectangular also the direction as often of this the is fastest also the verticaldirection growth of the rate fastest (Figure vertical 20). growth rate (Figure 20). TheThe pattern pattern on aon HP-template a HP-template usually usually consists consists of equal of inequal width in parallel width stripesparallel of stripes open for of growthopen for areasgrowth divided areas by divided interstitial by interstitial prohibited prohibited from growth from stripes. growth The stripes. patterns The are patterns oriented are in oriented two different in two directionsdifferent in directions a way that in will a way allow that growth will allow in the growth two opposite in the two crystallographic opposite crystallographic orientations oforientations interest. Theof ideainterest. is to The study idea independently is to study independently the propagation the behavior propagation of the behavior growing mesas,of the growing without theirmesas, mutualwithout interruption. their mutual Actually, interruption. the preparation Actually, andthe growthpreparation on the and HP growth template on isthe an HP important template step is an towardsimportant the fabricationstep towards of fully-patternedthe fabrication of (called fully-patterned orientation-patterned) (called orientation-patterned) templates. templates. FocusingFocusing on on the the crystal crystal cell cell of of a zinca zinc blended blended material, material, such such as as GaAs GaAs or or GaP, GaP, in in this this particular particular exampleexample (Figure (Figure 21), 21), one one can easilycan easily see that see in that the in z-direction the z-direction the structure the structure consists ofconsists alternating of alternating P- and Ga-layers.P- and Ga-layers. One also can One see also that can the twosee that mutually the two perpendicular mutually perpendicular crystallographic crystallographic planes CBB1C1 planesand ¯ DDCBB1A11Cwith1 and Miller DD1 indexesA1 with (1MillerI0) and indexes (110) determine (1Ī0) and (110) two oppositely determine oriented two oppositely directions oriented with Ga-atoms directions upwith on plane Ga-atoms CBB1 Cup1 and on plane Ga-atoms CBB down1C1 and on Ga-atoms plane DD down1A1. This on plane means DD to change1A1. This the means polarity to change we need the onlypolarity rotate thewe crystalneed only at 90 rotate degrees. the Atcrystal the same at 90 time, degrees. all bonds At the on thesame planes time, CBB all 1bondsC1 and on DD the1A 1planesare satisfied,CBB1C1 i.e., and there DD1 areA1 noarenet satisfied, charges i.e., or nothere elastic are force,no net which charges means or no itshould elastic beforce, easy which to cleave means the it wafershould in the be directions easy to cleave [1I0]¯ (which the wafer on a commercialin the directions GaP wafer [1Ī0] is(which perpendicular on a commercial to the major GaP flat) wafer and is [110],perpendicular which is perpendicular to the major toflat) the and minor [110], flat. which is perpendicular to the minor flat. TheThe patterns patterns related related to to the the result result shown shown in in Figure Figure 22 22were were deposited deposited on on one one “on-axis” “on-axis” (100) (100) GaP GaP wafer.wafer. One One should should bear bear in in mind, mind, however, however, thatthat aa miscutmiscut may also play play an an important important role role due due to to the theaforementioned aforementioned atomic atomic terraces terraces and and the the availability availability of of sites sites they they offer offer to to the the approaching approaching atoms. atoms. To Toexplore explore this option in anotheranother experimentexperiment wewe grewgrew simultaneously simultaneously on on one one “on-axis” “on-axis” (Figure (Figure 23 23a)a) ◦ andand on on one one with with 4 4°miscut miscut towards towards [111]B [111]B (100) (100) GaP GaP wafer wafer (Figure (Figure 23 b).23b).

FigureFigure 21. 21.Zinc-blended Zinc-blended structure structure of of GaP GaP with with crystallographic crystallographic orientations orientations that that provide provide opposite opposite crystalcrystal polarities. polarities. The The crystallographic crystallographic planes planes of interest of interest are outlined are outlined and shown and shown separately. separately. The right The imageright shows image a shows (100) GaP a (100) wafer GaP with wafer the with standard the standa directionsrd directions of the major of the and major minor and ﬂats. minor flats.

This also means that, if we deposit half-patterns in these two mutually perpendicular directions This also means that, if we deposit half-patterns in these two mutually perpendicular directions (Figure 22a), the domains growing on these patterns will have opposite crystallographic orientations (Figure 22a), the domains growing on these patterns will have opposite crystallographic orientations with nonlinear susceptibility with opposite signs. After the template was prepared in such a way, with nonlinear susceptibility with opposite signs. After the template was prepared in such a way, the the sample was placed in one of our HVPE reactors and thick growth was performed simultaneously sample was placed in one of our HVPE reactors and thick growth was performed simultaneously on on these oppositely oriented areas. The growths resulted in different vertical growth rates and different mesas’ shapes (Figure 22b). The faster vertical growth with rectangular mesas was

Crystals 2017, 7, 178 22 of 38

CrystalsCrystals 2017 2017, 7,, 7178, 178 2222 of of 38 38 these oppositely oriented areas. The growths resulted in different vertical growth rates and different mesas’achievedachieved shapes on on the (Figure the pattern pattern 22b). with with The stripes fasterstripes oriented verticaloriented growthalong along the the with [ 110[110 rectangular] ]direction, direction, mesaswhile while the was the growth achievedgrowth on on onthe the the patternstripesstripes with oriented oriented stripes along along oriented [110] [110] alongwas was about theabout [1 130% 030%] direction, slower slower and whileand resulted resulted the growth in in “triangular” “triangular” on the stripes or or “trapezoidal” oriented“trapezoidal” along [110]mesas’mesas’ was shape. aboutshape. 30% slower and resulted in “triangular” or “trapezoidal” mesas’ shape.

FigureFigureFigure 22. 22. 22.( a()a () Twoa Two) Two half-patterns half-patterns half-patterns depositeddeposited deposited in in two two two mu mutually mutuallytually perpendicular perpendicularperpendicular directions; directions; and and and (b ()b (bcross) )cross cross sectionsectionsection images images images of of theof the the layers layers layers growngrown grown onon on them them [63]. [63 [63].]. (R (Reproduced (Reproducedeproduced from fromfrom [63] [[63]63 with] with the the permission permission permission of of SPIE). of SPIE). SPIE).

FigureFigureFigure 23. 23. 23.(a ()a ( Growth)a Growth) Growth on on on one one one “on-axes”,“on-axes”, “on-axes”, and and ( b ())b one one) one misoriented misoriented misoriented with withwith 4° 44° ◦towards towards [111]B [111]B [111]B (100) (100) (100) GaP GaP GaP ◦ wafer;wafer;wafer; (c )(c Six-hour()c Six-hour) Six-hour thick thick thick HVPE HVPE HVPE growth growth growth on on on an an an HP-templateHP HP-template-template with with a patterna pattern deposited deposited deposited along along along [110][ 1[110]10] onon a a 4 misoriented4°4° misoriented misoriented towards towards towards [111]B [111]B [111]B (100) (100) (100) GaP GaP GaP wafer. wafer. wafer.

Crystals 2017, 7, 178 23 of 38 Crystals 2017, 7, 178 23 of 38 The trend, in point of view of growth rate and mesas shape, related to orientation [110] and [110] Theon both trend, wafers in point was of about view ofthe growth same. rateHowever, and mesas the growth shape, relatedrate on to the orientation “tilted” surface,[110] and as [110] we expected,on both wafers increased was additionally about the same. as in However, the [110] the case growth (Figure rate 23b: on the left “tilted” image) surface, for the as first we expected,time the [ ] topincreased layer surface additionally was, asin inaddition, the 110 hillock-frcase (Figureee. Thus 23b: thegrowth left image)on a pattern for the ﬁrstoriented time thealong top [110] layer surface was, in addition, hillock-free. Thus growth on a pattern oriented along [110] deposited on deposited on a 4° misoriented towards [111]B (100) GaP substrate turned out to be the most a 4◦ misoriented towards [111]B (100) GaP substrate turned out to be the most favorable case for favorable case for growth [58,63]. Growing for 6 h on such a HP template resulted in a 470 µm thick growth [58,63]. Growing for 6 h on such a HP template resulted in a 470 µm thick layer with excellent layer with excellent domain fidelity (Figure 23c). This proved the concept that GaP can be grown domain ﬁdelity (Figure 23c). This proved the concept that GaP can be grown successfully on patterned successfully on patterned templates and allowed us to take the next steps—preparation of OPGaP templates and allowed us to take the next steps—preparation of OPGaP templates and subsequent templates and subsequent thick HVPE growth on them. thick HVPE growth on them.

6.7.6.7. Preparation Preparation and and Growth Growth on on Orientation-Patt Orientation-Patternederned Templates Templates

6.7.1.6.7.1. Preparation Preparation of of OP OP Templates Templates TheThe main main reason reason for for researchers researchers [64] [64] to to attemp attemptt in in 2010 2010 heteroepitaxy heteroepitaxy of of OPZnSe OPZnSe on on OPGaAs OPGaAs templatestemplates was was not not (only) (only) the the small small lattice lattice mismatch mismatch between between ZnSe ZnSe and and GaAs GaAs but but rather rather the the lack lack of of nativenative OPZnSe OPZnSe templates. templates. OPGaAs OPGaAs templates templates at at that that time time were were the the only only available available OP OP templates templates and and thatthat is is why why in in 2009 2009 we we also also attempted attempted heteroepitaxy heteroepitaxy of of OPGaP OPGaP on on OPGaAs OPGaAs despite despite the the less less favorable favorable latticelattice match match between between GaP GaP and GaAs. The resultsresults publishedpublished in in 2011 2011 [ 57[57]] indicated indicated that that heteroepitaxy heteroepitaxy of ofOPGaP OPGaP on on OPGaAs, OPGaAs, if not if not great, great, is at is least at least possible—at possible—at about about 84 µm 84 domain µm domain width onewidth of theone domains of the domainsgrew at thegrew expense at the ofexpense its opposite of its neighboropposite neig buthbor after 1but hour after both 1 hour domains both were,domains all thewere, same, all the still same,presenting still presenting at the top layerat the surface top layer (Figure surface 24a). (Fig Beforeure 24a). becoming Before overgrown, becoming overgrown, the pattern survivedthe pattern for surviveda little while for a even littleat while domain even periodicity at domain as pe smallriodicity as 20 asµ smallm (Figure as 20 24 µmb). (Figure 24b).

FigureFigure 24. 24. FirstFirst attempts attempts on on HVPE HVPE heteroepitaxy heteroepitaxy of of OPGaP OPGaP on on OPGaAs OPGaAs templates templates at: at: (a (a) )larger larger and and ((bb)) finer ﬁner domain domain width. width. The The left left image image is is a a 3D 3D reco reconstructionnstruction showing showing the the to topp surfaces surfaces of of the the layer layer beforebefore and and after after the the growth growth and and its its cross cross section. section. The The right right image image is is a a cross cross section section of of another another layer layer growngrown at at the the same same conditions conditions on on a a pattern pattern with with finer ﬁner periodicity periodicity [57] [57].. (Reproduced (Reproduced from from [57] [57] with with the the permissionpermission of of EOS). EOS).

However,However, the the limited limited results results achieved achieved for for the much favorable ZnSe/GaAs case case [64] [64] convinced convinced everybodyeverybody in in the the field ﬁeld to to stop stop the the heteroepitaxial heteroepitaxial attempts attempts and and start start development development of of native native OPGaP OPGaP templatestemplates and and eventually OPGaPOPGaP homoepitaxy. homoepitaxy. The The idea idea was was painlessly painlessly to adoptto adopt these these techniques techniques from fromthe more the more mature mature OPGaAs OPGaAs template template preparation prepar processes—theation processes—the wafer bondingwafer bonding and the and MBE the assisted MBE assistedpolarity polarity alternation alternation process. process. However, However, due to due considerations to considerations presented presented in Section in Section 6.2.2 (high6.2.2 (high price, price,high EPD,high badEPD, wafer bad parallelism,wafer parallelism, lack of etchlack stop of etch material, stop etc.)material, this, practically,etc.) this, neverpractically, happened never to happenedthe extent to that the was extent expected. that was Some expected. amount Some of wafer amount bonded of wafer OPGaP bonded templates OPGaP (these templates templates (these are templatescheaper because are cheaper their fabricationbecause their does fabrication not require do MBEes not assistance) require MBE were assistance) produced anyway. were produced Later on anyway.the MBE Later technique, on the at MBE which technique, the polarity at which alternation the polarity of GaP alternation is achieved of by GaP the depositionis achieved of by a thinthe deposition of a thin non-polar Si-sublattice, was preferred, but wafer price and quality continued to

Crystals 2017, 7, 178 24 of 38 CrystalsCrystals 20172017,, 77,, 178178 2424 ofof 3838 non-polarbebe seriousserious Si-sublattice,obstacles.obstacles. ThickThick was HVPEHVPE preferred, homoepitaxyhomoepitaxy but wafer ofof price GaPGaP andwaswas qualityperformedperformed continued thenthen withwith to be variablevariable serious successsuccess obstacles. onon Thickbothboth templatetemplate HVPE homoepitaxy types.types. of GaP was performed then with variable success on both template types.

6.7.2.6.7.2. Homoepitaxy Homoepitaxy ofof OPGaPOPGaP onon OPGaPOPGaP TemplatesTemplates HomoepitaxyHomoepitaxy of of OPGaPOPGaP onon on OPGaPOPGaP OPGaP templates,templates, templates, usually,usually, usually, resultedresulted resulted inin in goodgood good domaindomain domain propagationpropagation propagation atat atgrowthgrowth growth durationsdurations durations lessless less thanthan than 33 hh 3 andand h and layerslayers layers nono thickerthicker no thicker thanthan than 60–12060–120 60–120 µmµm (Figureµ(Figurem (Figure 25a)25a) 25 [58].[58].a) [DueDue58]. to Dueto poorpoor to poortemplatetemplate template qualityquality quality andand andparasiticparasitic parasitic nucleationnucleation nucleation thatthat that becamebecame became moremore more pronouncedpronounced pronounced afterafter after thethe the 3rd3rd 3rd hour hour ofof growth,growth, however,however, thethe patternpattern on on layers layers thicker thicker than than 120–150 120–150 µmµµmm was was often often overgrown overgrown (Figure (Figure 25b).2525b).b). ThisThis processprocessprocess usually usuallyusually starts startsstarts with withwith roughening rougheningroughening of ofof the thethe layer layerlayer surface surfacesurface which, whwhich,ich, as asas it wasitit waswas already alreadyalready mentioned mentionedmentioned in Sectioninin SectionSection 6.4 6.4,6.4,, is is attackedis attackedattacked by byby the thethe additional additionaladditional amount amountamount of ofof HCl HClHCl released releasedreleased during duringduring the thethe parasitic parasiticparasitic nucleation nucleationnucleation ofof GaP.GaP. The TheThe V/III V/IIIV/III ratio ratio also also gradually gradually changes changes with with time time,time,, which which leads leads to to the the appe appearanceappearancearance ofof fastfast growinggrowing higherhigher MillerMillerMiller index indexindex facets facetsfacets tilted tiltedtilted towards towardstowards the thethe growth grgrowthowth direction—namely direction—namelydirection—namely these thesethese eventually eventuallyeventually overgrow overgrowovergrow the patternthethe patternpattern [69] [69][69] (Figure (Figure(Figure 25b). 25b).25b). More MoreMore details detailsdetails will willwill be further bebe furtherfurther presented. presented.presented. NextNext FigureFigure 2626 representsrepresents twotwo state-of-the-artstate-of-the-art experimentsexperimentsexperiments ofofof OPGaPOPGaPOPGaP onon OPGaPOPGaP templatestemplates withwith upup toto 150150 µµmµmm thickthickthick layerslayerslayers withwithwith goodgoodgood domaindomaindomain propagation,prpropagation,opagation, whosewhose thicknessthickness is is yetyet notnot enoughenough forfor frequencyfrequency conversionconversion testing. testing. This ThisThis probably probably sets sets the the limit limit of of homoepitaxyhomoepitaxy of of OPGaPOPGaP atat thethe appliedappliedapplied optimizedoptimized growthgrowthgrowth conditions. conditions.conditions. Thus ThusThus having, having,having, in inin addition, addition,addition, some somesome initial initialinitial experience experienceexperience (Figures (Figures(Figures 15 15and15 andand 24) 24)24) on GaP/GaAsonon GaP/GaAsGaP/GaAs we wewe already alreadyalready had hadhad enough enoughenough reasons reasonsreasons to steptoto stepstep into intointo the thethe county countycounty of heteroepitaxy.ofof heteroepitaxy.heteroepitaxy.

FigureFigure 25.25. ((aa)) One-hourOne-hourOne-hour homoepitaxialhomoepitaxial growthgrowth ofof OPGaPOPGaP onon OPGaP;OPGaP;OPGaP; ((bb)) AAA four-hourfour-hourfour-hour homoepitaxialhomoepitaxialhomoepitaxial growthgrowth indicatesindicates thatthat rougheningroughening of of thethe layerlayer surfacesusurfacerface andand overgrowthovergrowth ofof thethe patternpattern maymay occuroccur atat longerlonger experimentalexperimental timestimestimes [[63].[63].63]. (Reproduced(Reproduced fromfrom [[63][63]63] withwithwith thethethe permissionpermissionpermission ofofof SPIE).SPIE).SPIE).

Figure 26. (a,b) State-of-the-art experiments limit the good domain propagation to up to 150 µm FigureFigure 26.26. ((aa,,bb)) State-of-the-artState-of-the-art experimentsexperiments limitlimit thethe gogoodod domaindomain propagationpropagation toto upup toto 150150 µmµm thick OPGaP layer homoepitaxially grown on OPGaP templates [69]. (Reproduced from [69] with the thickthick OPGaPOPGaP layerlayer homoepitaxiallyhomoepitaxially growngrown onon OPGaPOPGaP templatestemplates [69].[69]. (Reproduced(Reproduced fromfrom [69][69] withwith thethe permission of SPIE). permissionpermission ofof SPIE).SPIE).

Crystals 2017, 7, 178 25 of 38 Crystals 2017, 7, 178 25 of 38

6.7.3. Heteroepitaxy of OPGaP on OPGaAs Templates 6.7.3.The Heteroepitaxy need of expensive of OPGaP growth on OPGaAs and characterization Templates tools and severe safety regulations restrict the Thegrowth need of of OPGaAs expensive to growth only a and few characterization places worldwide tools [60,70,71]. and severe Different safety regulations facilities restrictexpress thedifferent growth preferences of OPGaAs but to both only types, a few the places wafer worldwide fusion bonded [60,70 ,71[60,72,73]]. Different and the facilities MBE expressassisted differentpolarity preferencesinversion [74–76], but both OPGaAs types, thetemplates wafer fusion are currently bonded in [60 use.,72, 73Due] and to long the MBE extraordinary assisted polarity efforts inversionmm thick [ 74OPGaAs–76], OPGaAs with improved templates domain are currently fidelity in is use.not a Due surprise to long today extraordinary after a homoepitaxial efforts mm thickHVPE OPGaAs growth withof OPGaAs improved on OPGaAs domain templates fidelity is (Figure not a surprise 27a). This today led afterto a significant a homoepitaxial increase HVPE of the growthOPGaAs of conversion OPGaAs on efficiency OPGaAs templatesfrom 16% (Figure to 60% 27 (Figurea). This 27b). led toHomoepitaxy a significant increaseof OPGaAs of the can OPGaAs be now conversionconsidered efficiency a mature from process; 16% tohowever, 60% (Figure the 27bitteb).r Homoepitaxytruth is that ofOPGaAs OPGaAs has can not be nowyet reached considered the aproduction mature process; line. however, the bitter truth is that OPGaAs has not yet reached the production line. FollowingFollowing thethe OPGaAsOPGaAs developmentdevelopment stepssteps thethe pathpath ofof OPGaPOPGaP homoepitaxyhomoepitaxy cancan bebe alsoalsomarked marked withwith somesome successsuccess [77[77,78].,78]. AA relativelyrelatively goodgood domaindomain fidelityfidelity waswas achievedachieved andand asas aa resultresult frequencyfrequency conversionconversion ofof 16%16% atat 3.543.54µ µmm (Figure(Figure 28 28a)a) and,and, regretfully,regretfully, muchmuch lessless atat 11µ µmm waswas demonstrateddemonstrated [[49,50].49,50]. AtAt thethe samesame time, patterns patterns with with periods periods narrowe narrowerr than than 30–40 30–40 µmµm were were commonly commonly overgrown overgrown in the in theHVPE HVPE process process (Figure (Figure 28b), 28b), which which means means that that some some frequency frequency ranges ranges can can never be achievedachieved byby frequencyfrequency conversion conversion in in OPGaP. OPGaP. This, This, and notand forgetting not forgetting that the that main the advantage main advantage of GaP vs of GaAs—its GaP vs lowGaAs—its 2PA—is low namely 2PA—is in the namely 1–1.7 µ min rangethe 1–1.7 (Figure µm9b), range led us (Figure to believe 9b), that led improvements us to believe in thethat OPGaPimprovements material in quality the OPGaP are mandatory. material quality are mandatory.

FigureFigure 27.27. ((aa)) TheThe stepssteps takentaken towardstowards improvementsimprovements ofof thethe OPGaPOPGaP domaindomain fidelityfidelity (2004–2007)(2004–2007) precededpreceded the the following following increase increase of the of conversionthe conversion efficiency; efficiency; (b) The ( increaseb) The increase of the conversion of the conversion efficiency fromefficiency 16% to from about 16% 60% to was about achieved 60% was for aachieved 5-year period for a (2004–2009)5-year period [63 (2004–2009)]. (Reproduced [63]. from (Reproduced [63] with thefrom permission [63] with ofthe SPIE). permission of SPIE).

AtAt thatthat point, point, no no matter matter that that the the efforts efforts to to grow grow OPGaP OPGaP homoepitaxially homoepitaxially were were continuing continuing [52], [52], we startedwe started to realize to realize that the that low the quality, low quality, limited availabilitylimited availability and much and higher much price higher of GaP price compared of GaP tocompared GaAs (Table to GaAs2) would (Table always 2) would give always unavoidable give unavoidable obstacles thatobstacles finally that would finally not would allow not OPGaP allow homoepitaxyOPGaP homoepitaxy to make to it tomake the it production to the production line, exactly line, asexactly the case as the of OPGaAscase of OPGaAs which still which cannot still aftercannot so manyafter yearsso many of intensive years of research intensive efforts. resear However,ch efforts. our partialHowever, success our onpartial heteroepitaxy success ofon plainheteroepitaxy GaP on bare of plain GaAs GaP substrates on bare (Figures GaAs substrates 15 and 16 (Figures) and of 15 GaP and on 16) Si (Figureand of GaP 19) ledon usSi (Figure to believe 19) thatled us heteroepitaxy to believe that of OPGaP heteroepitaxy on OPGaAs of OPGaP is maybe on OPGaAs something is maybe that is worthsomething trying. that We is reconsideredworth trying. ourWe timidreconsidered 2009 attempts our timid (Figure 2009 24 )[attempts57] and (Figure this time 24 with) [57] more and confidencethis time with switched more from confidence homo- toswitched hetero-epitaxy. from homo- to hetero-epitaxy. ApplyingApplying ourour previousprevious experienceexperience andand anan optimizedoptimized setset ofof growthgrowth conditionsconditions inin anan improvedimproved reactorreactor configuration,configuration, wewe endedended upup withwith upup toto 300300 µµmm thickthick OPGaPOPGaP hetero-structureshetero-structures growngrown onon OPGaAsOPGaAs templates. templates. The The domain domain fidelity fidelity was was excellent excellent throughout throughout the whole the whole layer thicknesslayer thickness and along and thealong whole the samplewhole sample length (Figurelength (Figure 29) as the 29) resultsas the results were not were random not random but highly but repeatable.highly repeatable.

Crystals 2017, 7, 178 26 of 38 Crystals 2017,, 7,, 178178 26 of 38

Figure 28.28. ((a(a))) Frequency FrequencyFrequency conversion conversionconversion in in OPGaPin OPGaPOPGaP at 3.54 atat 3.543.54µm pumpµmµm pumppump beam beam [beam49]; ( b[49];[49];) Homoepitaxial ((b)) HomoepitaxialHomoepitaxial growth growthof OPGaP of onOPGaP a 3-inch on OPGaP a 3-inch template—at OPGaP template—at ﬁner pattern finer periods pattern the pattern periods is oftenthe pattern overgrown is often [50]. overgrown(Reproduced [50]. from (Reproduced [49,50] with from the permission [49,50] with of the SPIE). permission of SPIE).

Figure 29. A 3D-reconstruction of heteroepitaxyitaxy ofof OPGaPOPGaP onon OPGaAsOPGaAs templatestemplates basedbased onon microscopicmicroscopic Figure 29. A 3D-reconstruction of heteroepitaxy of OPGaP on OPGaAs templates based on microscopic imagesimages thatthat representrepresent thethe toptop surfacesurface andand thethe crosscross section—3section—3 imagesimages inin aa rowrow wherewhere thethe redred arrowsarrows showshow images that represent the top surface and the cross section—3 images in a row where the red arrows where they must be stitched [59]. (Reproduced from [59] with the permission of OSA). show where they must be stitched [59]. (Reproduced from [59] with the permission of OSA).

Regardless of the domain orientation, the domain top surfaces at shorter (2–3 h) experiments were Regardlessboth flat packed of the (Figure domain 30a) orientation, by a single the (100 domain)) facet.facet. top However,However, surfaces afterafter at shorter 4–84–8 hh ofof (2–3 growth,growth, h) experiments similarsimilar toto were both flat packed (Figure 30a) by a single (100) facet. However, after 4–8 h of growth, similar thethe OPGaAsOPGaAs case,case, thethe [110][110] oriented domain continued to be flat, while the opposite one (with to the OPGaAs case, the [110] oriented domain continued to be flat, while the opposite one (with orientation of the pattern along [110]) obtained a triangular shape—packed by two (111)pp facets;facets; oror orientation of the pattern along [110]) obtained a triangular shape—packed by two (111)p facets; or even a trapezoidal—packed by three facets, two (111)pp andand oneone (100)(100) facetfacet (Figure(Figure 30b).30b). TheThe reasonsreasons even a trapezoidal—packed by three facets, two (111)p and one (100) facet (Figure 30b). The reasons forfor thethethe differences differencesdifferences in inin the thethe growth growthgrowth rates ratesrates and theandand shapes thethe shapesshapes of the topofof thethe domain toptop domain surfacesdomain ofsurfacessurfaces the two ofof oppositely thethe twotwo oppositelyoriented domains oriented were domains discussed were earlier discussed [57,79 earlie]. Ther above[57,79]. results The above are from results growths are from on MBE growths assisted on MBEpolarity assisted inversion polarity OP templates, inversion whileOP templates, the results, while presented the results, in Figure presented 30c, are in from Figure growth 30c, are on waferfrom growthfused templates. on wafer They fused indicate templates. that further They refinements indicate thatthat of the furtherfurther etching refinementsrefinements and polishing ofof procedures thethe etchingetching during andand polishingthe template procedures fabrication during are still the necessary. template fabrication are still necessary.

Crystals 2017, 7, 178 27 of 38 Crystals 2017, 7, 178 27 of 38 Crystals 2017, 7, 178 27 of 38

Figure 30. Typical cross sections images of OPGaP grown by HVPE on Molecular Beam Epitaxy FigureFigure 30. Typical 30. Typical cross cross sections sections images images ofOPGaP of OPGaP grown grown by HVPEby HVPE on on Molecular Molecular Beam Beam Epitaxy Epitaxy (MBE) (MBE)(MBE) assisted assisted polarity polarity inversion inversion OPGaAsOPGaAs templatestemplates after: after: (a )( a2–3) 2–3 h ofh growth,of growth, and and(b) 6–8 (b )h 6–8of h of assisted polarity inversion OPGaAs templates after: (a) 2–3 h of growth, and (b) 6–8 h of growth; growth;growth; (c) (OPGaPc) OPGaP grown grown by by HVPE HVPE onon wafer bondedbonded OPGaAs OPGaAs templates templates [59]. [59]. (Reproduced (Reproduced from from (c) OPGaP grown by HVPE on wafer bonded OPGaAs templates [59]. (Reproduced from [59] with the [59] [59]with with the thepermission permission of of OSA). OSA). permission of OSA). 7. Discussion7. Discussion 7. Discussion 7.1. Impact7.1. Impact of the of theParasitic Parasitic Nu Nucleationcleation on on thethe HVPE Growth Growth 7.1. Impact of the Parasitic Nucleation on the HVPE Growth It seems that the next three points—the growth rate reduction (Figure 31a), the roughening of It seems that the next three points—the growth rate reduction (Figure 31a), the roughening of Itthe seems grown that layer the surface next (Figure three points—the 25b) and the growth overgrowth rate of reduction the pattern (Figure (Figures 31 25ba), theand roughening28b)—are of the grown layer surface (Figure 25b) and the overgrowth of the pattern (Figures 25b and 28b)—are the growndue to layerone and surface the same (Figure reason—parasitic 25b) and the reactions overgrowth that initiate of the and pattern promote (Figures parasitic 25 nucleationb and 28b)—are of due to one and the same reason—parasitic reactions that initiate and promote parasitic nucleation of due toGaP one on and the the nozzle same and reason—parasitic on the internal reactionswall of the that reactor initiate tube and in promotethe mixing parasitic zone before nucleation the of GaPchemicals on the nozzlereach the and growing on the substrate internal surface. wall Thes of thee reactions reactor weaken tube thein theprecursor mixing flows zone and before thus the GaP on the nozzle and on the internal wall of the reactor tube in the mixing zone before the chemicals chemicalsreduce reachthe local the saturation growing levelssubstrate of the surface. precursors Thes arounde reactions the substrate. weaken Al thethough precursor we already flows have and thus reach the growing substrate surface. These reactions weaken the precursor flows and thus reduce the reducehad the long local experience saturation (Figure levels 31a) of in the fighting precursors with lowering around the the grow substrate.th rate, Alwethough continue we to already believe have local saturation levels of the precursors around the substrate. Although we already have had long had thatlong this experience is a difficult (Figure task. For 31a) example, in fighting we have with discovered lowering that the increasing growth therate, growth we continue rate helps to up believe experience (Figure 31a) in fighting with lowering the growth rate, we continue to believe that this is a that tothis a certainis a difficult temperature task. For (726 example, °C) but after we havethat a discovered further temperature that increasing increase the does growth not make rate anyhelps up difficultsense, task. even For worse, example, it makes we the have average discovered layer ro thatughness increasing (RMS) start the growthgoing up rate again helps after up having to a certaina to a certain temperature (726 °C) but after that a further temperature increase does not make any temperatureminimum (726 at 726◦C) °C but(Figure after 30b). that a further temperature increase does not make any sense, even sense, even worse, it makes the average layer roughness (RMS) start going up again after having a worse, it makes the average layer roughness (RMS) start going up again after having a minimum at minimum at 726 °C (Figure 30b). 726 ◦C (Figure 30b).

Figure 31. (a) Growth rate as a function of the duration of growth on bare substrates, as well on HP and OP templates; (b) Progress in the growth rate increase (2010–2013) with one eye on the influence of the growth rate on the layer roughness; (c) Growth rate as a function of the substrate temperature FigureFigureat two 31. 31. ( differenta(a)) Growth Growth V/III rate rate ratios as as a[63]. a function function (Reproduce of of the thed from duration duration [63] with of of growth growththe permission on on bare bare of substrates, substrates,SPIE). as as well well on on HP HP andand OP OP templates; templates; ( b(b) Progress) Progress in in the the growth growth rate rate increase increase (2010–2013) (2010–2013) with with one one eye eye on on the the inﬂuence influence of theof the growthThe growth parasitic rate rate on nucleation theon the layer layer roughness;of GaProughness; is due (c) to Growth(c )the Growth same rate chemicalrate as aas function a functionreaction of the (Equationof the substrate substrate 6) temperaturethat temperature we would at twoloveat two differentto differenthave V/IIIon V/III the ratios ratiosgrowing [63 [63].]. (Reproduced layer (Reproduce surface. fromd fromBecause [63 [63]] with HClwith the isthe permission a permissionby-product of SPIE).of in SPIE). this reaction, this “unexpected” HCl amount, once released up front of the substrate, starts to attack, etching back and The The parasiticparasitic nucleation nucleation of of GaP GaP is isdue due to the to thesame same chemical chemical reaction reaction (Equation (Equation 6) that 6) we that would we wouldlove to love have to on have the on growing the growing layer layersurface. surface. Because Because HCl is HCl a by-product is a by-product in this in thisreaction, reaction, this this“unexpected” “unexpected” HCl HCl amount, amount, once once released released up front up front of the of thesubstrate, substrate, starts starts to attack, to attack, etching etching back back and

Crystals 2017, 7, 178 28 of 38 and thus roughening the already grown layer surface. At the same time, the involvement of equal amounts of P and Ga atoms in these parasitic reactions leads to a gradual change in the initially establishedCrystals 2017 V/III, 7, 178 ratio. Thus, looking at Figure 30c, where we studied the growth rate as28 a of function 38 of the growth temperature at two different V/III ratio (2.25 and 4.41), one can see that at the higher V/IIIthus ratio roughening after a certain the already temperature grown (714 layer◦C) surface. the growth At the rate same starts time, to the decrease involvement even withof equal a further amounts of P and Ga atoms in these parasitic reactions leads to a gradual change in the initially increase of the growth temperature. Thus an uncontrollable self-increase of the V/III ratio during established V/III ratio. Thus, looking at Figure 30c, where we studied the growth rate as a function of the experimentthe growth temperature may not only at two slow different down V/III the ratio growth (2.25 but and also 4.41), “eject” one can the see process that at the out higher of the V/III growth range.ratio Reducing after a certain the growth temperature rate, however, (714 °C) isthe only growth one rate of the starts impacts to decrease of the even V/III with ratio. a further Changes of this ratioincrease in oneof the or growth the other temperature. direction Thus can an promote uncontrollable conditions self-inc forrease the appearanceof the V/III ratio of fasterduring growingthe phosphorusexperiment or galliummay not only terminated slow down facets the tiltedgrowth towards but also the“eject” growing the process direction, out of suchthe growth as (111) range. or (311), that canReducing overgrow the growth the pattern. rate, however, is only one of the impacts of the V/III ratio. Changes of this ratio Thein one parasitic or the nucleation other direction can be can suppressed promoteby cond introducingitions for intothe theappearance reactor an of additional faster growing peripheral HCl flowphosphorus to prevent or gallium the appearance terminated offacets GaP tilted crystallites towards onthe thegrowing quartz direction, surfaces such or, as if they(111) haveor (311), already appeared,that can to overgrow etch them the back pattern. or at least to prevent their further enlargement. However, this additional The parasitic nucleation can be suppressed by introducing into the reactor an additional peripheral HCl may also etch back the already grown GaP layer, which means a very delicate balance peripheral HCl flow to prevent the appearance of GaP crystallites on the quartz surfaces or, if they is necessary.have already appeared, to etch them back or at least to prevent their further enlargement. However, Itthis is additional worth mentioning peripheral HCl again may that also etch the influenceback the already of parasitic grown GaP nucleation layer, which on means growing a very surface morphologydelicate balance is less is pronounced necessary. during heteroepitaxy compared to homoepitaxy. Obviously, the higher surfaceIt is worth quality mentioning of the GaAs again wafers, that the relative influenc toe the of qualityparasitic of nucleati the GaPon wafers, on growing and thesurface resulting highermorphology quality of is the less fabricated pronounced OPGaAs during templates heteroepitaxy provide compared conditions to homoepitaxy. for a better Obviously, start of the the HVPE growth.higher As surface a consequence, quality of the GaAs surface wafers, of the relative growing to the layer quality has of a the higher GaP qualitywafers, and and the thus, resulting is not that complianthigher toquality the HCl of the attacks, fabricated so theOPGaAs growth templates can continue provide uninterruptedconditions for a forbetter a longerstart of periodthe HVPE of time. One shouldgrowth. bearAs a inconsequence, mind that HClthe surface can attack of the not growin onlyg thelayer existing has a higher defects quality but also and createthus, is new not that etch-pits, compliant to the HCl attacks, so the growth can continue uninterrupted for a longer period of time. i.e., new spots for nucleation on the wafer surface that could be called “secondary” nucleation points. One should bear in mind that HCl can attack not only the existing defects but also create new This secondaryetch-pits, i.e., nucleation new spots could for nucleation be another on reason the wafer for earlier surface aggravation that could ofbe thecalled layer “secondary” quality. Thenucleation next two points. proposed This secondary approaches nucleation suggest could rather be to another surround reason the for problems earlier aggravation related to the of parasiticthe nucleationlayer quality. than to solve them: The next two proposed approaches suggest rather to surround the problems related to the 1. To produce thicker OPGaP layers some crystal growers (Figure 28b) interrupt the growth after parasitic nucleation than to solve them: several hours, clean the reactor tube, slightly polish, if necessary, the already grown layer surface, 1.and To perform produce a secondthicker OPGaP growth layers on the some same crystal sample growers to gain (Figure some 28b) more interrupt thickness. the growth after several hours, clean the reactor tube, slightly polish, if necessary, the already grown layer 2. Instead of losing years in optimizing growth condition and reactor configuration the layer surface, and perform a second growth on the same sample to gain some more thickness. thickness (optical aperture) can be increased by precise face-to-face bonding at higher temperature 2. Instead of losing years in optimizing growth condition and reactor configuration the layer for twothickness device-sized (optical aperture) pieces cut can from be oneincreased already by grownprecise OP-sample.face-to-face bonding For this atpurpose higher the triangulartemperature parts for of thetwo topdevice-sized domain surfacespieces cut (Figure from one32a), already if there grown are any, OP-sample. must be For polished this off first.purpose Than the triangular pieces must parts be of facedthe top to domain each other surfaces as attempts(Figure 32a), to if align there the are sameany, must orientation be domainspolished must off befirst. made, Than althoughthe pieces must from be the faced optical to each point other of as view attempts this istonot align that the criticalsame for quasi-phase-matchingorientation domains (Figuremust be 32 made,b,c). Thenalthough the piecesfrom the should optical be point pressed of view and keptthis is in not this that way for severalcritical hours for quasi-phase-matching at higher temperature (Figure to allow 32b,c). mutual Then diffusion.the pieces should be pressed and kept in this way for several hours at higher temperature to allow mutual diffusion.

FigureFigure 32. Two 32. Two pieces pieces of grownof grown OPGaAs: OPGaAs:( (aa)) before,before, and and (b (b) )after after polishing polishing off offthe thetriangular triangular parts parts on on the domainthe domain top surfaces;top surfaces; (c) two(c) two already already faced faced against against each each other other andand alignedaligned OPGaP OPGaP pieces pieces ready ready for highfor temperature high temperature bonding. bonding.

Crystals 2017, 7, 178 29 of 38

Whether the defective bond interface area will have an unacceptably high absorption during the wave mixing process is still under investigation. However, the important thing is that, if this works, the face-to-face bonding procedure can be repeated again and again to increase the optical aperture up to 5–10 mm and more.

7.2. Some Considerations on Heteroepitaxy of OP Nonlinear Optical Materials In contrast to homoepitaxy, the growth mechanisms of heteroepitaxy of many HVPE growing wideband compound semiconductor materials are unclear yet, even for traditional growth orientations [80]—not to mention the case, when they are orientation-patterned materials. Focusing on the particular GaP/GaAs case, we believe that there are at least three critical questions that must be addressed. First, what is the advantage of GaAs as an alternative substrate for HVPE growth of GaP in comparison to its own native GaP substrate? Second, how in any way can hundreds of microns thick GaP be grown on GaAs substrates at such a large lattice mismatch of −3.6%, while growths such as ZnSe/GaAs at the negligible lattice mismatch of only +0.3% (Figure 13b) [64] ended up with only limited success (Figure 13c)? Third, what is the influence of the periodic polarity alternation on the growth and the material quality? Focusing on the first question, we should bear in mind, as already mentioned, once a 3D-island has been formed, due to the Ehrlich–Schwoebel (ES) effect [4] it will continue to grow three-dimensionally [4]. Namely, the ES barrier facilitates the 3D-growth by repulsing the atoms from the terrace that are trying to cross the step to the next lower level. Thus the ES barrier increases the supersaturation on the terrace, allowing the nucleation of a second, a third, etc., layer on the top of the terrace, preventing flattening out of the 3D-island. Recalling that the etch pits, associated with ends of screw dislocations, promote spiral growth and 3D formation, one can easily see why a crystal surface with a higher EPD (GaP) will result after the growth in a rougher surface morphology than a surface with a lower EPD (GaAs). Fortunately, this also turned out to be true in the case of GaP/GaAs heteroepitaxy (Figure 11). Although the ES barrier, when calculated for some metals, is relatively small and plays a negligible role at higher temperatures, it is still widely used in modeling and analysis of many morphological transformations that occur on the growing crystal surface [81,82], as the aim is to ensure good roughness control. In the case of semiconductors some sources [83] led us to believe that the ES effect may play a more tangible role even at close-to-equilibrium processes such as HVPE. Unfortunately, we have not yet determined any range of growth parameters for GaAs and GaP where ES barriers are relevant. As for the second question, we should bear in mind that in heteroepitaxy there is a drastic difference in the film growth mechanisms depending on the sign of the lattice mismatch, i.e., depending on whether the film is growing under compressed or tensile stress. The reason is the anharmonicity of the interatomic potentials; the repulsive branch is much steeper than the attractive branch. The mismatch (or misfit) between GaP and GaAs is negative, as far as the GaP crystal cell is smaller. Respectively, the growing layer is under tensile strain. In this case, i.e., when a material with a lattice constant b0 (GaP) grows on another with a lattice constant a0 (GaAs) and a0 > b0, misfit dislocations (MDs) that represent unsaturated (dangling) bonds, will appear originating from the material with the smaller lattice parameter (GaP) with a periodicity that is proportional to the difference a0 − b0 [84]. When this occurs, according to the misfit dislocations concept, depends on the relation between the forces that keep in place the atoms of the substrate or the atoms of the growing layer ΨAA and ΨBB, and the interfacial force ΨAB [4]. In the case of GaP/GaAs ΨAB >> ΨBB and ∼ ΨAB = ΨAA, which means that, initially, the growing crystal B (GaP) is homogeneously strained to fit to the substrate crystal A (GaAs), i.e., the interfacial force ΨAB is strong enough to produce a pseudomorphous growth. Thus, the appearance of the expected misfit dislocations will be postponed for about 10–20 atomic monolayers at the expense of an accommodating and linearly increasing elastic strain [4] (Figure 33). However, beyond some critical thickness the pseudomorphous growth will become energetically unfavorable and the homogeneous strain will be released in a misfit dislocation. Crystals 2017, 7, 178 30 of 38

Crystals 2017, 7, 178 30 of 38

Aslead a result, to an thealmost periodic perfect distortion match inof the both crystal lattices planes will inlead some to areas, an almost separated perfect by matchstripes,in where the crystal the planestwo lattices in some are areas, out of separated registry. by stripes, where the two lattices are out of registry.

Figure 33. Pseudomorphous growth in the case of heteroepitaxy of PbSe on PbS [4]. (Reproduced Figure 33. Pseudomorphous growth in the case of heteroepitaxy of PbSe on PbS [4]. (Reproduced from [4] with some modifications with the permission of World Scientific).The small image of the from [4] with some modifications with the permission of World Scientific).The small image of the lower lower right corner shows a TEM image of an area of the pseudomorphous growth during right corner shows a TEM image of an area of the pseudomorphous growth during heteroepitaxy of heteroepitaxy of GaP on GaAs [85]. (Reproduced from [85] with the permission of AIP Publishing). GaP on GaAs [85]. (Reproduced from [85] with the permission of AIP Publishing). The key to answering the second question is that it has been experimentally confirmed [86] that tensileThe keyfilms to grow answering pseudomorphically the second question up to a ismuch that greater it has been thickness experimentally compared confirmedwith compressed [86] that tensileones at films the growsame absolute pseudomorphically value of the lattice up to mismat a muchch greater (misfit). thickness In other words, compared the tensile with stress compressed (the onesnegative at the mismatch) same absolute favors value planar of the growth, lattice while mismatch the compressed (misfit). In stress other (the words, positive the tensile misfit) stress favors (the negativethe formation mismatch) of 3D favors islands. planar That growth, is why a while small the negative compressed misfit stressis preferable (the positive [87] in misfit)the case favors of theheteroepitaxy. formation of Indeed, 3D islands. most Thatattempts is why to demons a smalltrate negative Stranski–Krastanov misfit is preferable growth [87 are] in based the case on of heteroepitaxy.material combinations Indeed, most that attempts provide to demonstrategrowth at a Stranski–Krastanovpositive misfit (Ge/Si, growth InAs/GaAs) are based onin materialorder combinationspurposely to thatpromote provide 3D growth. growth In at our a positivecase, compar misfiting (Ge/Si, the layer InAs/GaAs) quality in the in two order opposite purposely cases, to promotegrowth 3Dof GaP/GaAs growth. In and our GaAs/GaP case, comparing, we did the find layer differences quality in in growth the two rate, opposite surface cases, morphology growth of GaP/GaAsand crystalline and GaAs/GaP, quality in wefavor did of find the differencesGaP/GaAs ingrowth. growth A rate, deeper surface study morphology would probably and crystalline reveal qualitydifferences in favor in ofthe the dislocation GaP/GaAs core growth. energies, A deeper the gliding study velocities, would probably and the reveal nucleation differences barriers. in the However, at this point it is still too early to tell with 100% confidence that growth under tensile dislocation core energies, the gliding velocities, and the nucleation barriers. However, at this point strain is more favorable in the case of GaP/GaAs. All of the above is of course valid for thin film it is still too early to tell with 100% confidence that growth under tensile strain is more favorable in growth. In our case of thick film growth, as was already mentioned, the misfit dislocations should be the case of GaP/GaAs. All of the above is of course valid for thin film growth. In our case of thick introduced beyond some critical thickness (the thickness of the pseudomorphous growth) [88] and film growth, as was already mentioned, the misfit dislocations should be introduced beyond some then one should again have the usual crystal growth. However, having a good quality growth at the criticalbeginning, thickness i.e., a (the good thickness start obviously of the pseudomorphous plays an important growth) role in the [88 ]later and stages then one of growth. should again have the usualTo crystaladdress growth. the third However, question havingabout the a good role qualityof the periodic growth atpolarity the beginning, alternation, i.e., we a goodshould start obviouslyrecall that plays the dangling an important bonds, role depending in the later on stagesthe surf oface growth. polarity of the adjacent crystal planes, can actTo as addresseither donors the third or acceptors. question As about such, the they role may of the constitute periodic deep polarity energy alternation, levels in the we energy should gap recall thatand the thus dangling play the bonds, role of depending recombination on thecenters. surface This polarity means that of the a higher adjacent dangling crystal bond planes, density can can act as eithersignificantly donors orchange acceptors. the material As such, properties—electrical they may constitute and deep optical. energy Alternating levels in the periodically energy gap the and thuscrystal play polarity the role of of an recombination OP material, statistically, centers. This will means strive that to equalize a higher the dangling number bond of donors density and can significantlyacceptors (Figure change the34). material At the properties—electricalsame time, each act andof polarity optical. alternation Alternating changes periodically entirely the crystalthe polarityenvironment of an OP around material, the neighboring statistically, ions will at strivethe inte torface. equalize This means the number a change of of donors the probability and acceptors for (Figuresome 34dangling). At the bonds same time,to be each satu actrated, of polarityi.e., to disappear—most alternation changes probably entirely unequally the environment for opposite around theorientations. neighboring Thus, ions by at reducing the interface. the number This of means dangling a change bonds, ofthe the periodic probability polarity for alternation some dangling may bondshave to even be saturated,a “healing” i.e., effect to disappear—most expressed, first, by probably improvement unequally of the for optical opposite quality orientations. (due to reduced Thus, by reducingnumber the of donors number and of danglingacceptors) bonds, and, second, the periodic by making polarity the bond alternation between may substrate have even and agrowing “healing” effectlayer expressed, stronger. first, by improvement of the optical quality (due to reduced number of donors and acceptors) and, second, by making the bond between substrate and growing layer stronger.

Crystals 2017, 7, 178 31 of 38 Crystals 2017, 7, 178 31 of 38 Crystals 2017, 7, 178 31 of 38

Figure 34. Heteroepitaxy of OPGaP on OPGaAs templates. For simplicity the pseudomorphous Figure 34.34. HeteroepitaxyHeteroepitaxy of of OPGaP OPGaP on OPGaAson OPGaAs templates. templates. Forsimplicity For simplicity the pseudomorphous the pseudomorphous growth growth is not presented on the image [4]. ((Reproduced from [4] with some modifications with the growthis not presented is not presented on the image on the [4 ].image ((Reproduced [4]. ((Reprod fromuced [4] with from some [4] with modifications some modifications with the permission with the permission of World Scientific). permissionof World Scientific). of World Scientific). The thickness of the pseudomorphous growth and the periodicity of the misfit dislocations, Thedetermined thickness for aof particular the pseudomorphous case of heteroepitaxy, growth growth for an andexampled the periodicity for GaP/GaAs, of of maythe the misfit misfitplay the dislocations, role of determinedcriteria byfor which a particular to determine case which of heteroepitaxy, other cases of forfor heteroepitaxy exampleexample forfor would GaP/GaAs,GaP/GaAs, be also favorable. may may play play the role of criteria by To which find out to determineon which crystallographic which other cases plane of the heteroepitaxy misfit dislocations wouldwould may bebe appear alsoalso favorable.favorable. we should take Toanother find find outlook out on onat whichthe which zinc-blended crystallographic crystallographic GaP structure plane plane the (F theigure misfit misfit 21), dislocations focusing dislocations this may time may appear on appear the we diagonal should we should take anothertake anothercrystallographic look lookat the at plane thezinc-blended zinc-blended (111) (ADC GaP1) (Figure GaP structure structure 35a). (Figure (Figure 21), 21 ),focusing focusing this this time time on on the the diagonal crystallographic plane (111) (ADC1)) (Figure (Figure 35a).35a).

Figure 35. (a) The zinc-blended GaP structure. The crystallographic plane of interest ADC1 is outlined; (b) conducting the tensile biaxial strain σ during GaP/GaAs heteroepitaxy [85]. (Reproduced from [85] with the permission of AIP Publishing).

Figure 35.35. (a(a)) The The zinc-blended zinc-blended GaP GaP structure. structure. The crystallographicThe crystallographic plane ofplane interest of interest ADC1 is ADC outlined;1 is The biaxial strain σ along the [100], [010], and [001] axes can be resolved into a uniaxial shear (b) conducting the tensile biaxial strain σ during GaP/GaAs heteroepitaxy [85]. (Reproduced from [85] outlined;stress τ on ( bthe) conducting {111} dislocation the tensileglade planebiaxial [85], strain as shown σ during on Figure GaP/GaAs 35b by heteroepitaxythe red arrows. [85]. As with the permission of AIP Publishing). (Reproducedindicated, the from arrow [85] directions with the permission are related of to AIP the Publishing). case of growth at a negative misfit when the growing layer is under tensile stress. In the opposite case (positive misfit and the layer growing Theunder biaxial compressive strainstrainσ σstrain)along along the the the arrow [100], [100], directions, [010], [010], and and of [001] co [001]urse, axes axesmust can can bebe resolvedopposite be resolved [89,90]. into ainto uniaxial The a signuniaxial shear of the shear stress stress τ on the {111} dislocation glade plane [85], as shown on Figure 35b by the red arrows. As τ onthe {111} dislocation glade plane [85], as shown on Figure 35b by the red arrows. As indicated, indicated,the arrow directionsthe arrow aredirections related toare the related case of to growth the case at aof negative growth misﬁt at a whennegative the misfit growing when layer the is growingunder tensile layer stress. is under In thetensile opposite stress. case In (positivethe opposite misﬁt case and (positive the layer misfit growing and underthe layer compressive growing under compressive strain) the arrow directions, of course, must be opposite [89,90]. The sign of the

Crystals 2017, 7, 178 32 of 38 Crystals 2017, 7, 178 32 of 38

strain)misfit determines the arrow directions, not only the of course,type of must the strain be opposite but also [89 how,90]. Thethe strain sign of will the be misfit relaxed. determines In the notindicated only the case type (negative of the strain mismatch; but also tensile how thelayer strain growth) willbe the relaxed. strain Inwill the be indicated released caseas plastic (negative via mismatch;dislocation tensilenucleation layer and growth) glide theand strain form willtwo-dimensional be released as (2D) plastic layers via [80,91]. dislocation That nucleation is why plastic and gliderelaxation and formis encouraged two-dimensional if the goal (2D) is to layers grow [ 80me,91tamorphic]. That is buffers why plastic [92]. In relaxation contrast, is when encouraged the strain if theis elastic goal is(positive to grow misfit; metamorphic compressive buffers layer [92 growth)]. In contrast, there will when be thean elastic strain relaxation is elastic (positive that will misfit;result compressivein surface roughening, layer growth) which there can will be used be an to elastic drive relaxationquantum dot that self-assembly will result in surfaceprocesses roughening, [80,93,94]. whichAll this can clearly be used explains to drive why quantum the HVPE dot deposition self-assembly of GaP processes on (001) [80 GaAs,93,94 results]. All this in forming clearly explainssmooth, whyplanar the layers HVPE [59,69], deposition while of the GaP elastic on (001) relief GaAs on both results (110) in GaAs forming and smooth, (111)A planarGaAs layersleads to [59 a, 693D], whileself-assembly the elastic of reliefGaP oninto both dislocation-free (110) GaAs and nanostructures (111)A GaAs [90]. leads Resolving to a 3D self-assembly the biaxial strain of GaP σ intointo dislocation-freeuniaxial shear strain nanostructures τ on the [{111}90]. Resolvingdislocation the glide biaxial plane strain [85,95]σ into will uniaxial make shearthe atoms strain onτ on {111} the {111}experience dislocation a shear glide force, plane which [85,95 will] will provoke make the an atoms irreversible on {111} translocation experience a from shear their force, equilibrium which will provokepositions an and, irreversible thus, will translocation gives a rise fromto formation their equilibrium of dislocations positions (Figure and, 36a) thus, and will stacking-faults gives a rise to formation(Figure 36b). of dislocationsIn this process (Figure on 36e a)should and stacking-faultsnot be surprised, (Figure if 36theb). expected In this process 60° dislocation one should is notdissociated be surprised, into a if pair the expectedof a 30° 60and◦ dislocation a 90° Shockley is dissociated partial dislocations, into a pair of as a far 30◦ asand the a 90sum◦ Shockley of their partialformation dislocations, energies is as about far as 2/3 the of sum theof formation their formation energy energiesneeded for is about the 60° 2/3 dislocation. of the formation In addition energy to neededthat, there for is the an 60 explicitly◦ dislocation. defined In additionsequence to in that, which there the is two an explicitly partial dislocations defined sequence must pass—one in which thewill two lead partial and the dislocations other will must follow pass—one which, willagain, lead strongly and the depends other will on follow the sign which, of again,misfit, stronglytype of dependsstrain, and on thesubstrate sign of misfit,orientation. type ofFinally, strain, andthe substrate30° and orientation.90° partials Finally, can be the bound 30◦ and by 90 a◦ partialsnarrow canstacking-fault be bound by“ribbon” a narrow (Figure stacking-fault 36b). It is “ribbon”worth mentioning (Figure 36 b).that It al isthough worth mentioningthe deposition that techniques although thein the deposition compared techniques articles [59,69,85,90] in the compared are quite articles different [59,69, 85(HVPE,90] are and quite MBE) different the obtained (HVPE andresults MBE) on theGaP/GaAs obtained are results similar. on GaP/GaAs are similar.

◦ ◦ Figure 36.36. (a) Formation of 3030° and 9090° Shockley partialpartial dislocations;dislocations; (b)) aa TEMTEM imageimage ofof thethe twotwo partialspartials bondedbonded byby aa narrow stacking-faultstacking-fault “ribbon”“ribbon” [[85].85]. (Reproduced fromfrom [[85]85] withwith thethe permissionpermission of AIPAIP Publishing).Publishing).

After allall thesethese considerations considerations the the expected expected scenario scenario is: theis: ﬁrstthe severalfirst several (10–20) (10–20) monoatomic monoatomic layers oflayers pseudomorphous of pseudomorphous growth shouldgrowthform should a high form quality a high dislocation-free quality dislocation-free area [85,88 area] (see [85,88] also Figure (see also 33), followedFigure 33), by followed a several byµ ma several thick area µm whichthick area should which be heavilyshould be populated heavily populated with dislocations with dislocations that are a resultthat are of a the result release of the of therelease strain of builtthe strain during buil thet during pseudomorphous the pseudomorphous growth. With growth. thickness With increasethickness a reductionincrease a ofreduction the dislocation of the dislocation density is expected. density is Such expected. “healing Such effect” “healing is observed effect” inis manyobserved thick in grown many materials,thick grown for materials, example infor HVPE example GaN. in Finally, HVPE theGaN. top Finally, layer surface the top should layer besurface with smoothshould be surface with morphologysmooth surface populated morphology with the populated typical features with the for 2Dtypi growth.cal features Indeed, for after 2D agrowth. rough looking Indeed, interface after a whichrough islooking probably interface due to which somethermal is probably decomposition due to some of thermal the GaAs decomposition substrate our TEMof the and GaAs SEM substrate studies showedour TEM a and roughly SEM 2–3 studiesµm thick showed area a heavilyroughly populated 2–3 µm thick with area dislocations heavily populated as expected with [85 ]dislocations in the (111) zone—screwas expected [85] + mixed in the when (111) G zone—screw = 002 and edge + mi +xed mixed when when G = G 002 = 02 and2 (Figure edge +37 mixeda–d). Withwhen thickness G = 022 increase(Figure 37a–d). the dislocation With thickness density graduallyincrease the reduces, dislocation exactly density as it happens gradually in GaNreduces, (Figure exactly 37e) as [96 it]. Ithappens turned outin thatGaN this (Figure area was 37e) highly [96]. transparentIt turned (95%out that transmittance) this area forwas the highly wavelength transparent of the pump(95% beamtransmittance) that we intended for the to usewavelength for quasi-phase-matching, of the pump which beam was that what we we wanted.intended Finally, to use the GaPfor layerquasi-phase-matching, is ﬁnished, as we expectedwhich was [90 ],what with we an extremelywanted. Finally, smooth the top surfaceGaP layer (RMS is withinfinished, 1 nm as inwe a 1expectedµm × 1 µ[90],m AFM with image) an extremely with the smooth typical featurestop surface for 2D(RMS growth within (Figure 1 nm 37 inf). a 1 µm × 1 µm AFM image) with the typical features for 2D growth (Figure 37f).

Crystals 2017, 7, 178 33 of 38 Crystals 2017, 7, 178 33 of 38

Figure 37.37. ((aa––dd)) TEM TEM cross cross section section images images with with different different magnification magniﬁcation in in the area near to thethe GaP/GaAs interface;interface; (e(e)) a a cross cross section section image image of a thickof a thick HVPE HVPE grown grown GaN layer GaN that layer shows that a reductionshows a ofreduction the dislocation of the dislocation density with density layer thicknesswith layer [ 96thickness]. (Reproduced [96]. (Reproduced from [96] withfrom the [96] permission with the ofpermission AIP Publishing); of AIP Publishing); (f) top surface (f) top of surface thick HVPE of thick grown HVPE GaP grow onn GaAs GaP on substrate GaAs substrate showing showing smooth surfacesmooth morphology.surface morphology.

8. Conclusions History shows that delay and failures in sciencescience and technology are often related not only to gaps and lapses in thethe materialsmaterials sciencescience but alsoalso toto miscommunicationmiscommunication between different scientiﬁcscientific communities. Focusing Focusing on on laser laser so sourcesurces development, as as an example, we could tell that without frequency conversion (including via QPM) this develo developmentpment would not be as far as it is now. In the late 90s 90s planar planar technology technology was was already already advanced advanced en enoughough to topropose propose techniques techniques for forfabrication fabrication of OP of OPtemplates, templates, so soon so soon thick thick OPGaAs OPGaAs structures structures we werere grown grown on on such such templates, templates, which immediately replaced the current stacksstacks ofof GaAs wafers [[97].97]. Thus rapid progress was achieved in about a decade (Figure(Figure 2727).). Then “suddenly” the same scientists and engineers realized that there was a number of excellent, convenient and readily available pump laserlaser sources in the 1–1.7 µmµm range that regretfully could not be used with GaAs due to itsits strongstrong 2PA2PA inin thethe samesame frequencyfrequency range.range. Then the world turned with eyes full with hope to GaP, oneone ofof thethe mostmost promisingpromising alternatives.alternatives. Compared to GaAs, GaP looked wonderful, havinghaving aboutabout thethe samesame nonlinearnonlinear susceptibility, susceptibility, but but much much lower lower 2PA 2PA (Figure (Figure9 and9 and Table Table1). 1). The The problem problem was was that that native native OPGaP OPGaP templates templates were were not not yet yet available, available, plus plus to adopt to adopt 1:1 the1:1 OPGaAsthe OPGaAs template template preparation preparation technique technique for fabrication for fabrication of OPGaP of OPGaP templates templates did not did sound not an sound easy task.an easy Meanwhile, task. Meanwhile, two timid two heteroepitaxial timid heteroepitaxia attemptsl wereattempts conducted were conducted on what people on what had people available had at theavailable time, OPGaAsat the time, templates—HVPE OPGaAs templates—HVPE growth of OPGaP growth on of OPGaAs OPGaP [57 on] andOPGaAs a relevant [57] and PVT a (Physical relevant VaporPVT (Physical Transport) Vapor growth Transport) of OPZnSe growth on of OPGaAs OPZnSe [on64]. OPGaAs Although, [64]. in Although, comparison in comparison to the negligible to the latticenegligible mismatch lattice betweenmismatch ZnSe between and GaAs,ZnSe and the latticeGaAs, mismatchthe lattice between mismatch GaP between and GaAs GaP looks and GaAs huge (Figurelooks huge 13b), (Figure the results 13b), [57 the,58 results] (Figures [57,58] 15 and (Figures 24) were 15 notand less24) discouragingwere not less (ondiscouraging the contrary) (on than the thosecontrary) obtained than those in the obtained more favorable in the more case [favorabl64] (Figuree case 13c). [64] Nobody (Figure at13c). this Nobody point, however, at this point, paid anyhowever, attention paid to any the factattention that theto twothe fact groups, that [57the,64 two] used groups, quite [57,64] different used growth quite techniques, different growth HVPE andtechniques, PVT. Otherwise, HVPE and somebody PVT. Otherwise, would have somebody realized that would HVPE, have as anrealized easier-to-control that HVPE, technique, as an couldeasier-to-control probably be technique, attempted could with probably more success be attempted for growth with of OPZnSemore success on OPGaAs for growth templates. of OPZnSe The authorson OPGaAs [57] templates. made, indeed The anotherauthors really[57] made, discouraging indeed another attempt really this timediscouraging (Figure 13 attempta) to grow this time GaP on(Figure Si (with 13a) the to pointgrow thatGaP theon latticeSi (with mismatch the point between that the GaP lattice and mismatch Si is negligible—see between GaP Figure and 13Si b),is whichnegligible—see resulted onlyFigure in 13b), a few which small resulted GaP crystallites only in a randomly few small distribute GaP crystallites on the randomly Si substrate distribute surface on the Si substrate surface (Figure 13a). Then the word “heteroepitaxy” was definitely forgotten and the community rushed-off at brisk pace towards development of techniques for preparation of

Crystals 2017, 7, 178 34 of 38

(Figure 13a). Then the word “heteroepitaxy” was definitely forgotten and the community rushed-off at brisk pace towards development of techniques for preparation of native OPGaP templates and homoepitaxy. Thus several more years were lost before the research parties interested in this could figure out that OPGaP would be another cruel battle on the muddy scientific field: GaP was growing with a lower growth rate than GaAs, as the process was accompanied by severe parasitic reactions that further reduced the growth rate and deteriorated the layer quality. At the same time, it turned out that the GaP market is quite narrow as the two-inch GaP wafers offered are 7–8 times more expensive than the related GaAs wafers. Moreover, in point of view of EPD and wafer parallelism (Table2), their quality was incomparably lower. At such a low quality, practically, nobody could expect either a good quality of the prepared OPGaP templates or a good subsequent thick HVPE growth on such low quality templates. Thus, in a moment, the old idea [57] for heteroepitaxy of OPGaP on OPGaAs started to look attractive again, even more, it turned out that after optimizing the reactor configuration and the applied growth conditions it worked fairly well [59,69] (Figures 29 and 30). Once having success with the less favorable case, GaP/GaAs, of heteroepitaxy [59,69,98–100], we tried to analyze again why we could not grow GaP on Si. We recalled that a combination between two growth techniques, one close-to-equilibrium and one far from it, is, actually, a daily routine in the growth of III-Nitrides, for example, thick HVPE growth of GaN on a thin AlN layer deposited in advance by MOCVD on sapphire [101]. So, applying the same approach to the HVPE growth of GaP on a thin MOCVD GaP/Si although it did not sound like discovering America, it was however quite successful (Figure 19b). Then, little by little, we started to realize with a great sigh of relief that there will be more combinations of materials and more growth and template preparation approaches, i.e., there will be from now on more Americas to be discovered. In this exciting journey heteroepitaxy will continue invisibly to do its “little” great favors to the world of Optics and Electronics.

Acknowledgments: I would like to express my most cordial gratitude to all quoted authors as well to my devoted coworkers. I am also very thankful to the AFOSR for their longstanding interest and support of our projects under contract 13RY09COR. Conﬂicts of Interest: The author declares no conﬂict of interest.

References

1. John Vincent Atanasoff Biography. Available online: http://jva.cs.iastate.edu/jvabio.php (accessed on 18 April 2017). 2. Raspberry PI Zero: The 5$ Computer. Available online: http://www.raspberry-pi-zero (accessed on 12 June 2017). 3. Who Invented the Transistor? Available online: http:/www.computerhistory.org/atchm/who-invented-the- transistor/ (accessed on 18 April 2017). 4. Markov, I.V. Crystal Growth for Beginners: Fundamentals of Nucleation, Crystal Growth and Epitaxy, 2nd ed.; World Scientific: Singapore; New York, NY, USA, 2003; p. 43. 5. Wang, L.G.; Kratzer, P.; Schefller, M.; Moll, N. Formation and stability of self-assembled coherent islands in highly mismatched heteroepitaxy. Phys. Rev. Lett. 1999, 82, 4042–4045. [CrossRef] 6. Brehm, M.; Montalenti, F.; Grydlik, M.; Vastola, G.; Lichtenberger, H.; Hrauda, N.; Beck, M.J.; Fromherz, T.; Schäffler, F.; Miglio, L.; et al. Key role of the wetting layer in revealing the hidden path of Ge/Si(001) Stranski-Krastanow growth onset. Phys. Rev. 2009, B80, 20FF5321. [CrossRef] 7. Stranski, I.; Krastanov, L. Über die kristallisation von alkalihalogenidkristallen auf fluorit. Akad. Wiss. Lit. Mainz 1939, 146, 797. 8. Bauer, E. Phänomenologische Theorie der Kristallabscheidung an Oberflächen. Kristallografiya 1958, 110, 372. [CrossRef] 9. Wang, X.H.; Jia, H.Q.; Guo, L.W.; Xing, Z.G.; Wang, Y.; Pei, X.J.; Zhou, J.M.; Chen, H. White light-emitting diodes based on a single InGaN emission layer. Appl. Phys. Lett. 2007, 91, 161912. [CrossRef] 10. Bhattacharyya, A.; Moustakas, T.D.; Zhou, L.; Smith, D.J.; Hug, W. Deep ultraviolet emitting AlGaN quantum wells with high internal quantum efficiency. Appl. Phys. Lett. 2009, 94, 181907. [CrossRef] Crystals 2017, 7, 178 35 of 38

11. Zhang, H.; Chen, Y.; Zhou, G.; Tang, G.; Wang, Z. Wetting layer evolution and its temperature dependence during self-assembly of InAs/GaAs quantum dots. Nanoscale Res. Lett. 2012.[CrossRef][PubMed] 12. Liu, L.; Edgar, J.H. Substrates for gallium nitride epitaxy. Mater. Sci. Eng. 2002, 37, 61–127. [CrossRef] 13. Hellman, E.S. Properties, Processing and Applications of Gallium Nitride and Related Semiconductors: Alternative Oxide Substrates for GaN Heteroepitaxy; IEE, INSPEC: London, UK, 1999. 14. Trampert, A.; Brandt, O.; Ploog, K.H. Crystal structure of Group III Nitrides. Semicond. Semimet. 1997, 50, 167–192. [CrossRef] 15. Amano, H. Metalorganic vapor phase epitaxial growth of a high quality GaN film using an AlN buffer layer. Appl. Phys. Lett. 1986, 48, 353–355. [CrossRef] 16. Amano, H.; Takeuchi, T.; Yamaguchi, S.; Wetzel, C.; Akasaki, I. Characterization of the crystalline quality on GaN on sapphire and ternary allows. Electr. Commun. Jpn. Part 2 1998, 81, 65–71. 17. Nakamura, S. GaN growth using GaN buffer layer. Jpn. J. Appl. Phys. 1991, 30, L1705–L1707. [CrossRef] 18. Park, H.; Chan, H.M. A novel process for the generation of pristine sapphire surfaces. Thin Solid Films 2002, 422, 135–140. [CrossRef] 19. Park, H.; Chan, H.M.; Vinci, R.P. Patterning of sapphire substrates via a solid state conversion process. J. Mater. Res. 2005, 20, 417–423. [CrossRef] 20. Falub, C.V.; Kanel, H.; Isa, F.; Bergamashini, R.; Marzegalli, A.; Chrastina, D. Scaling hetero-epitaxy from layers to three-dimensional crystals. Science 2012, 335, 1330–1334. [CrossRef][PubMed] 21. Isa, F.; Salvalaglio, M.; Dasilva, Y.; Meduna, M.; Barget, M.; Jung, A.; Kreliger, T.; Isella, G.; Emi, R.; Pezzoli, F.; et al. Highly mismatched, dislocation-free SiGe/Si heterostructures. Adv. Mater. 2016, 28, 884–888. [CrossRef] [PubMed] 22. Zubia, D.; Hersee, S.D. Nanoheteroepitaxy: The application of nonostructuring and substrate compliance to the heteroepitaxy of mismatched semiconductor materials. J. Appl. Phys. 1999, 85, 6492–6496. [CrossRef] 23. Lester, S.D.; Ponce, F.A.; Craford, M.G.; Steigerwald, D.A. High dislocation densities in high efficiency GaN-based light-emitting diodes. Appl. Phys. Lett. 1995, 66, 1249–1251. [CrossRef] 24. Ponce, F.A.; Bour, D.P. Nitride-based semiconductors for blue and green light-emitting devices. Nature 1997, 386, 351–359. [CrossRef] 25. Moustakas, T.D. The role of extended defects on the performance of optoelectronic devices in nitride semiconductors. Phys. Status Solidi 2013, 210, 169–174. [CrossRef] 26. Zytkiewicz, Z.R. Laterally overgrown structures as substrates for lattice mismatched epitaxy. Thin Solid Films 2002, 412, 64–75. [CrossRef] 27. Tadatomo, K.; Okagawa, H.; Ohuchi, Y.; Tsunekawa, T.; Imada, Y.; Kato, M.; Taguchi, T. High output power InGaN ultraviolet light-emitting diodes fabricated on patterned substrates using metalorganic vapor phase epitaxy. Jpn. J. Appl. Phys. 2001, 40, L583. [CrossRef] 28. Wang, H.; Zhou, S.; Lin, Z.; Hong, X.; Li, G. Enhanced light emitting diode light extraction efficiency by an optimized spherical cap-shaped patterned sapphire substrates. Jpn. J. Appl. Phys. 2013, 52, 092101. [CrossRef] 29. Davis, R.F.; Bishop, S.M.; Mita, S.; Collazo, R.; Reitmeier, Z.J.; Sitar, Z. Epitaxial growth of gallium nitride. AIP Conf. Proc. 2007, 916, 520. [CrossRef] 30. Zheleva, T.S.; Ashmawy, W.M.; Jones, K.A. Pendeo-Epitaxy versus Lateral Epitaxial Overgrowth of GaN: A Comparative Study via Finite Element Analysis. Phys. Status Solidi 1999, 176, 545–551. [CrossRef] 31. Davis, R.F.; Gehrke, T.; Linthicum, K.J.; Zheleva, T.S.; Preble, E.A.; Rajagopal, P.; Zorman, C.A.; Mehregany, M. Pendeo-epitaxial growth of thin films of gallium nitride and related materials and their characterization. J. Cryst. Growth 2001, 225, 134–140. [CrossRef] 32. Detchprohm, T.; Hiramatsu, K.; Itoh, K.; Akasaki, I. Relaxation Process of the Thermal Strain in the

GaN/α-Al2O3 Heterostructure and Determination of the Intrinsic Lattice Constants of GaN Free from the Strain. Jpn. J. Appl. Phys. 1992, 31, L1454. [CrossRef] 33. Tsuchiy, H.; Hasegawa, F.; Okukura, H.; Yoshida, S. Comparison of Hydride Vapor Phase Epitaxy of GaN layers on cubic GaN/(100) GaAs and hexagonal Ga/(111) GaAs substrates. Jpn. J. Appl. Phys. 1994, 33, 6448–6453. [CrossRef] 34. List of Laser Types. Available online: https://en-wikipedia.org/wiki/List_of_laser_types (accessed on 12 June 2017). 35. Maiman, T. Stimulated Radiation in Ruby. Nature 1960, 187, 493–494. [CrossRef] Crystals 2017, 7, 178 36 of 38

36. Tacke, M. Lead Laser Sources. Philos. Trans. R. Soc. Lond. 2001, A359, 547–566. [CrossRef] 37. Ruby Laser. Available online: https://en-wikipedia.org/wiki/Ruby_laser (accessed on 12 June 2017). 38. Infrared Lasers. Available online: https://en.wikipedia.org/wiki/Infrared_lasers (accessed on 14 June 2017). 39. Kildaland, H.; Mikkelsen, J.C. The nonlinear coefﬁcient phase matching and optical damage in the

chalcopyrite AgGaSe2. Opt. Commun. 1973, 9, 315–318. [CrossRef] 40. Myers, L.E.; Bosenberg, W.R. Periodically poled lithium niobate and quasi-phase-matched optical parametric oscillators. IEEE J. Quantum Electron. 1997, 33, 1663–1672. [CrossRef] 41. Wu, Y.C.S.; Feigelson, R.S.; Route, R.K.; Zheng, D.; Gordon, L.A.; Fejer, M.M.; Byer, R.L. Low Optical Loss Wafer Bonded GaAs Structures for Quasi-Phase-Matched Second Harmonic Generation. Mater. Res. Soc. Symp. Proc. 1998, 484, 481–485. [CrossRef] 42. Grisard, A.; Lallier, E.; Gérard, B. Quasi-phase-matched gallium arsenide for versatile mid-infrared frequency conversion. Opt. Mater. Express 2012, 2, 1021. [CrossRef] 43. Schunemann, P.G.; Zetzler, S.D. Future directions in quasi-phasematched semiconductors for mid-infrared lasers. Proc. SPIE 2011, 7917, 79171F-1. [CrossRef] 44. Hurlbut, W.; Lee, Y.; Vodopyanov, K.; Kuo, P.; Fejer, M. Multi-photon absorption and nonlinear refraction of GaAs in the mid-infrared. Opt. Lett. 2007, 32, 668–670. [CrossRef][PubMed] 45. Skauli, T.; Kuo, P.S.; Vodopyanov, K.L.; Pinguet, T.J.; Levi, O.; Eyres, L.A.; Harris, J.S.; Fejer, M.M.; Gerard, B.; Becouarn, L.; et al. Improved dispersion relations for GaAs and applications to nonlinear optics. J. Appl. Phys. 2003, 94, 6447. [CrossRef]

46. Madarasz, F.L.; Dimmock, J.O.; Dietz, N.; Bachmann, K.J. Sellmeier parameters for ZnGaP2, ZnGaP2 and GaP. J. Appl. Phys. 2000, 87, 1564. [CrossRef] 47. Tomita, I. Fabrication and characterization of a quasi-phase-matched GaP optical device for terahertz-wave generation. Opt. Mater. 2009, 32, 323–328. [CrossRef] 48. Matsushita, T.; Ohta, I.; Kondo, T. Quasi-Phase-Matched Parametric Fluorescence in a Periodically Inverted GaP Waveguide. Appl. Phys. Exp. 2009, 2, 0611011–0611013. [CrossRef] 49. Schunemann, P.G.; Pomeranz, L.A.; Magarrell, D.J. Optical parametric oscillation in quasi-phase-matched GaP. Proc. SPIE 2015, 9347.[CrossRef] 50. Pomeranz, L.A.; Schunemann, P.G.; Magarrell, D.J.; McCarthy, J.C.; Zawilski, K.T.; Zelmon, D.E. 1-µm-pumped OPO based on orientation-patterned GaP. Proc. SPIE 2015, 9347.[CrossRef] 51. Peterson, R.; Bliss, D.; Lynch, C.; Tomich, D. Progress in orientation-patterned GaAs for next-generation nonlinear optical devices. Proc. SPIE 2008, 6875.[CrossRef] 52. Insero, G.; Clivati, C.; D’Ambrosio, D.; De Natale, P.; Santambrogio, S.; Schunemann, P.G.; Zondy, J.J.; Borri, S. Difference frequency generation in the mid-infrared with orientation-patterned gallium phosphide crystals. Opt. Lett. 2016, 41, 5114–5117. [CrossRef][PubMed] 53. Levine, B.F.; Bethea, C.G. Nonlinear Susceptibility of GaP: Relative Measurement and Use of Measured Values to Determine a Better Absolute Value. Appl. Phys. Lett. 1972, 20, 272. [CrossRef] 54. Hu, X.P.; Xu, P.; Zhu, S.N. Engineered quasi-phase-matching for laser techniques. Photonics Res. 2013, 1, 171–185. [CrossRef] 55. Spitzer, W.G.; Gershenzon, M.; Frosch, C.J.; Gibbs, D.F. Optical absorption in n-type gallium phosphide. J. Phys. Chem. Sol. 1959, 11, 339–341. [CrossRef] 56. Yasakov, D.A.; Pikhtin, A.N.; Ulyanov, V.I. Optical properties of gallium phosphide grown by ﬂoating zone. Mater. Res. Bull. 1969, 4, 839–848. [CrossRef] 57. Tassev, V.; Bliss, D.; Snure, M.; Bryant, G.; Peterson, R.; Bedford, R.; Yapp, C.; Goodhue, W.; Termkoa, K. HVPE growth and characterization of GaP on different substrates and patterned templates for frequency conversion devices. J. Eur. Opt. Soc. 2011, 6, 110171–110177. [CrossRef] 58. Tassev, V.; Snure, M.; Peterson, R.; Schepler, K.L.; Bedford, R.; Manna, J.M.; Vangala, S.; Goodhue, W.; Lin, A.; Harris, J.; et al. Progress in orientation-patterned GaP for next-generation nonlinear optical devices. Proc. SPIE 2013, 8604.[CrossRef] 59. Tassev, V.L.; Vangala, S.R.; Peterson, R.D.; Kimani, M.M.; Snure, M.; Stites, R.W.; Guha, S.; Slagle, J.E.; Ensley, T.R.; Syed, A.A.; et al. Heteroepitaxial growth of OPGaP on OPGaAs for frequency conversion in the IR and THz. Opt. Mater. Express 2016, 6, 1724–1737. [CrossRef]

60. Lallier, E.; Brevignon, M.; Leho, J. Efﬁcient second-harmonic generation of a CO2 laser with a quasi-phase-matched GaAs crystal. Opt. Lett. 1998, 23, 1511–1513. [CrossRef][PubMed] Crystals 2017, 7, 178 37 of 38

61. Lin, A.; Fenner, D.B.; Termkoa, K.; Allen, M.; Moulton, P.; Lynch, C.; Bliss, D.F.; Goodhue, W.D. Wafer bonding of GaP, Wafer-fused orientation-patterned GaAs. Proc. SPIE 2008, 6875.[CrossRef] 62. Kuzmenko, P.J.; Little, S.L.; Ikeda, Y.; Kobayashi, N. Fabrication and testing of diamond-machined gratings in ZnSe, GaP, and bismuth germanate for the near infrared and visible. Proc. SPIE 2008, 7018.[CrossRef] 63. Tassev, V.; Snure, M.; Vangala, S.; Kimani, M.; Peterson, R.; Schunemann, P. Growth and study of nonlinear optical materials for frequency conversion devices with applications in defense and security. Proc. SPIE 2014, 9253.[CrossRef] 64. Singh, N.B.; Kanner, G.S.; Berghmans, A.; Kahler, D.; Lin, A.; Wagner, B.; Kelley, S.P.; Knuteson, D.J.; Holmstrom, R.; Schepler, K.L.; et al. Characteristics of thick ZnSe films on quasi-phase-matched (QPM) GaAs substrates. J. Cryst. Growth 2010, 312, 1142–1145. [CrossRef] 65. III-V Semiconductors. Available online: html://www.tf.uni-kiel.de/matwis/amat/semitech_en/kap_2/ backbone/r2_3_1.html (accessed on 12 June 2017). 66. Gil, E.; André, Y.; Cadoret, R.; Trassoudaine, A. Handbook of Crystal Growth: Hydride Vapor Phase Epitaxy for Current III–V and Nitride Semiconductor Compound Issues; Elsevier: Amsterdam, The Netherlands, 2015; pp. 51–93. 67. Derenge, M.A.; Zheleva, T.S.; Jones, K.A.; Shah, P.B.; Ewing, D.; Molstad, J.; Lee, U.; Ervin, M.H.; Stepp, D.N. Pendeo-Epitaxy Process Optimization of GaN for Novel Devices Applications; ARL-TR-4426; Army Research Lab.: Adelphi, MD, USA, 2008; pp. 20783–21197. 68. Gil-Lafon, E.; Napierala, J.; Castelluci, D.; Pimpinelli, A.; Cadoret, R.; Gerard, B. Selective growth of GaAs by HVPE: Keys for accurate control of the growth morphologies. J. Cryst. Growth 2001, 222, 482–496. [CrossRef] 69. Tassev, V.L.; Vangala, S.; Peterson, R.; Kimani, M.; Snure, M.; Markov, I. Homo and heteroepitaxial growth and study of orientation-patterned GaP for nonlinear frequency conversion devices. Proc. SPIE 2016, 9731. [CrossRef] 70. Bliss, D.F.; Lynch, C.; Weyburne, D.; O’Hearn, K.; Bailey, J.S. Epitaxial growth of thick GaAs on orientation-patterned wafers for nonlinear optical applications. J. Cryst. Growth 2006, 287, 673–678. [CrossRef] 71. Eyres, L.A.; Tourreau, P.J.; Pinguet, T.J.; Ebert, C.B.; Harris, J.S.; Fejer, M.M.; Gerald, B.; Lallier, E. All-epitaxial fabrication of thick, orientation-patterned GaAs films for nonlinear optical frequency conversion. Appl. Phys. Lett. 2001, 79, 904–906. [CrossRef] 72. Peterson, R.; Whelan, D.; Bliss, D.; Lynch, C. Improved material quality and OPO performance in orientation-patterned GaAs. Proc. SPIE 2009, 7197.[CrossRef] 73. Faye, D.; Grisard, A.; Lallier, E.; Gérard, B.; Kieleck, C.; Hirth, A. Orientation-patterned gallium arsenide: Engineered materials for infrared sources. SPIE Newsroom 2008.[CrossRef] 74. Ebert, C.B.; Eyres, L.A.; Fejer, M.M.; Harris, J.S. MBE growth of antiphase GaAs films using GaAs/Ge/GaAs heteroepitaxy. J. Cryst. Growth 1999, 201, 187. [CrossRef] 75. Schunemann, P.G.; Setzler, S.D.; Mohnkern, L.; Pollak, T.M.; Bliss, D.F.; Weyburne, D.; O’Hearn, K. 2.05-µm-Laser-Pumped Orientation-Patterned Gallium Arsenide (OPGaAs) OPO. In Proceedings of the Conference on Lasers and Electro-Optics 2005, Baltimore, MD, USA, 22 May 2005. 76. Lynch, C.; Bliss, D.F.; Zens, T.; Lin, A.; Harris, J.S.; Kuo, P.S.; Fejer, M.M. Growth of mm-thick orientation-patterned GaAs for IR and THZ generation. J. Cryst. Growth 2008, 310, 5241–5247. [CrossRef] 77. Tassev, V.; Snure, M.; Peterson, R.; Bedford, R.; Bliss, D.; Bryant, G.; Mann, J.M.; Goodhue, W.; Vangala, S.; Termkoa, K.; et al. Epitaxial growth of quasi-phase-matched GaP for nonlinear applications: Systematic process improvements. J. Cryst. Growth 2012, 352, 72–77. [CrossRef] 78. Schunemann, P.G.; Mohnkern, L.; Vera, A.; Yang, X.S.; Lin, A.C.; Harris, J.S.; Tassev, V.; Snure, M. All-Epitaxial Growth of Orientation-Patterned Gallium Phosphide (OPGaP). In Proceedings of the Advances in Optical Materials 2012, San Diego, CA, USA, 1–3 February 2012. 79. Vangala, S.R.; Tassev, V.L.; Kimani, M.; Snure, M.; Peterson, R. Development of Thick Orientation Patterned GaP for Frequency Conversion in the Mid IR and THz Region. In Proceedings of the 2014 39th International Conference on Infrared, Millimeter, and Terahertz waves (IRMMW-THz), Tucson, AZ, USA, 14–19 September 2014. 80. Joyce, B.A.; Vvedensky, D.D. Self-organized growth on GaAs surfaces. Mater. Sci. Eng. 2014, 46, 127–176. [CrossRef] Crystals 2017, 7, 178 38 of 38

81. Pierre-Louis, O.; D’Orsogna, M.R.; Einstein, T.L. Edge Diffusion during Growth: The Kink Ehrlich-Schwoebel Effect and Resulting Instabilities. Phys. Rev. Lett. 1999, 82, 3661–3664. [CrossRef] 82. Leal, F.F.; Ferreira, S.O.; Ferreira, S.C. Modelling of epitaxial film growth with an Ehrlich-Schwoebel barrier dependent on the step height. J. Phys. Condense Matter 2011, 23, 292201. [CrossRef][PubMed] 83. Ayers, J.E.; Kujofsa, T.; Rago, P.; Raphael, J.E. Heteroepitaxy of Semiconductors: Theory, Growth and Characterization, 2nd ed.; CRC Press: Boca Raton, FL, USA, 2017; pp. 154–165. 84. Holt, D.B. Misfit dislocations in semiconductors. J. Phys. Chem. Solids 1966, 27, 1053–1067. [CrossRef] 85. Simmonds, P.J.; Lee, M.L. Tensile-strained growth on low-index GaAs. Appl. Phys. Lett. 2011, 99, 123111. [CrossRef] 86. Markov, I.; Milchev, A. The effect of anharmonicity in epitaxial interfaces: Equilibrium structure of thin epitaxial films. Surf. Sci. 1984, 136, 519–531. [CrossRef] 87. Zangwill, A. Physics at Surfaces; Cambridge University Press: New York, NY, USA, 1988. 88. Walther, M.; Biehl, M.; Kinzel, Y. Formation and consequences of misfit dislocations in heteroepitaxial growth. Phys. Status Solidi 1997, 4, 3210–3220. [CrossRef] 89. Simmonds, P.J.; Lee, M.L. Self-assembly on (111)-oriented III-V surfaces. Appl. Phys. Lett. 2010, 97, 153101. [CrossRef] 90. Simmonds, P.J.; Lee, M.L. Tensile GaAs (111) quantum dashes with tunable luminescence below the bulk bandgap. J. Appl. Phys. 2012, 112, 054313. [CrossRef] 91. Pachinger, D.; Groiss, H.; Lichtenberger, H.; Stangl, J.; Hesser, G.; Schaffler, F. Stranski-Krastanov growth of tensile strained Si islands on Ge (001). Appl. Phys. Lett. 2007, 91, 233106. [CrossRef] 92. Simon, J.; Tomasulo, S.; Simmonds, P.J.; Romero, M.; Lee, M.L. Metamorphic GaAsP buffers for growth of wide-bandgap InGaP solar cells. J. Appl. Phys. 2011, 109, 013708. [CrossRef] 93. Leonard, D.; Krishnamurthy, M.; Reaves, C.M.; Denbaars, S.P.; Petroff, P.M. Direct formation of quantum-sized dots from uniform coherent islands of InGaAs on GaAs surfaces. Appl. Phys. Lett. 1993, 63, 3203. [CrossRef] 94. Eaglesham, D.J.; Cerullo, M. Dislocation-free Stranski-Krastanov growth of Ge on Si (100). Phys. Rev. Lett. 1990, 64, 1943. [CrossRef][PubMed] 95. Kvam, E.P.; Hull, R. Surface orientation and stacking fault generation in strained epitaxial growth. J. Appl. Phys. 1993, 73, 7407. [CrossRef] 96. Sheu, J.K.; Loe, M.L.; Lai, W.C. Effect of low-temperature grown GaP layer on reduced leakage current of GaN Schottky diodes. Appl. Phys. Lett. 2005, 86, 052103. [CrossRef] 97. Szilagyi, A.; Hordvik, A.; Schlossberg, H. A quasi-phase-matching technique for efficient optical mixing and frequency doubling. J. Appl. Phys. 1976, 47, 2025–2932. [CrossRef] 98. Vangala, S.; Peterson, R.; Snure, M.; Tassev, V. Development of orientation-patterned GaP grown on foreign substrates for QPM frequency conversion devices. Proc. SPIE 2017, 10088.[CrossRef] 99. Vangala, S.; Kimani, M.; Peterson, R.; Sites, R.; Snure, M.; Tassev, V. Thick orientation-patterned growth of GaP on wafer-fused GaAs templates by hydride vapor phase epitaxy for frequency conversion. Opt. Mater. 2016, 60, 62–66. [CrossRef] 100. Tassev, V.L.; Peterson, R.D. Heteroepitaxial Growth of Orientation-Patterned Materials on Orientation- Patterned Substrates. U.S. Patent 9,647,156, 9 May 2017. 101. Paskova, T.; Tungasmita, S.; Valcheva, E.; Svedberg, E.B.; Arnaudov, B.; Evtimova, S.; Persson, P.A.; Henry, A.; Beccard, R.; Heuken, M.; et al. Hydride Vapour Phase Homoepitaxial Growth of GaN on MOCVD-Grown ‘Templates’. Mater. Res. Soc. Int. J. Nitride Semicond. Res. 2000, 5, 131–137. [CrossRef]

© 2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).