A Brief Introduction to Laser Diodes

Transcription

1 A Brief Itroductio to Laser Diodes This defiitely wo't do for a course, but if you're ot familiar with laser diodes, this might be a good lace to start. I am deliberately light o the equatios ad details i the hoe that it will be easier to gras a uderstadig of this stuff. If you're goig to secialize further, the you will eed to dig further ad get the details! Diodes First, let's just start with lai diodes. A diode is simly made from 2 ieces of semicoductor material (or likely, oe iece with 2 differet regios). These are usually called the doed regio ad the doed regio. Hece, a diode is commoly simly referred to as a juctio. Why? Because, ormally a iece of semicoductor (e.g. Silico) usually does't have ay free electros, so it is a oor coductor (hece semicoductor). It is ot as bad as a ure isulator, or as good as a good coductor, such as metals. I order to rovide free electros, what you do is "doe" the semicoductor. By doig, you mix i some imurities ito the semicoductor (i this case, Silico). Just have a look at the table of elemets ad go o either side of Silico, say Phoshorous (P) ad Boro (B), for examle. The extra P atoms try to fit ito the Silico lattice, but oos! it's got a extra electro! So i order to fit i, it maages to "give u" oe to be a somewhat free electro. Similarly for the B, it tries to fit ito the Silico as well, but it's missig a electro comared to its eighbours, so it will have a extra "hole", which it will try to fill i, so the hole ca have the ossibility of gettig shuffled aroud. You ca't ut i too much, of course, or else it is't Silico ay loger ad the whole lattice gets messed u... So, let's go iside ad take a look at what haes. I a itrisic (udoed) iece of semicoductor, it has a certai badga betwee the valece ad coductio bad, ad Fermi eergy level which usually lies i betwee. This is described by the followig equatio: where: E g = badga eergy k B = Boltzma costat T = temerature m h * = effective hole mass m e * = effective electro mass [Equatio 1] (Note that the effective mass is't the same as the actual mass, eve assumig that holes have mass!). The badga diagram is borig. Flat lie kid of borig, so we'll ot bother lookig at that. Now, whe we add doats, we chage thigs a bit, ad i articular, we refer to the Fermi level eergy (or robably more aroriately, the quasifermi level, but we'll use it the same way here). The quasifermi level i a doed material ca be described by: [Equatio 2] [Equatio 3] 2002/02/15 10:21:27 Page 1 of 7

2 where: E f = quasifermi level i the doed material E f = quasifermi level i the doed material = carrier cocetratio of electros = carrier cocetratio of holes i = itrisic carrier cocetratio (Note: ad are usually equal to the doig cocetratio). You ca see from the equatios the that the Fermi level for the regio goes u, ad the regio goes dow relative to itrisic material. Now, whe you joi the two regios, the quasifermi level still has to lie u somehow. It ca't move. So what haes the is the valece ad coductio bad, while fixed relative to the Fermi level i their ow regio, shifts to comesate. This we show i the diagram below. E f E c E v Figure 1: Eergy bad diagram for juctio Pretty cool, eh? But othig haes yet. Come o, you ca't have currets flowig from othig. Mother Nature adjusts everythig accordigly so that there is't ay et curret flow with a iece of juctio just sittig aroud. But, we ca do somethig about it. If we aly a ositive voltage to with resect to (that is, oly has to be of a higher voltage tha ), this is called forward bias. What you ca imagie haeig is that the bad diagram gets "squished", the ad regios get closer together (vertically). E f E c V alied E f E v Figure 2: Eergy bad diagram for juctio uder forward bias Now somethig iterestig haes. All of a sudde the excess electros i the regio see a lower barrier to the side, ad a buch of them go tumblig over. Ad similarly for the holes. Ad that is curret flow! That i itself is't that iterestig, I mea you could do that with a iece of wire. But cosider the oosite case, reversebias, where is of a higher voltage tha. The istead of gettig closer together, the bads shift lower ad the bads shift higher. There is ow a bigger barrier for the excess electros o the side to climb over if they wat to get over to the side. There is ractically o curret flow. Now that becomes more iterestig, because you have a juctio which effectively coducts oe way, but ot the other. 2002/02/15 10:21:27 Page 2 of 7

3 Light Emittig Diodes We'll talk about regular Light Emittig Diodes (LED's) first, as laser diodes are merely secial case of LED's. Now i the case of a light emittig diode, what you wat is the carriers i the middle regio to liger log eough to recombie ad emit light. What is haeig is a hole becomes available, ad the electro says "hey, I ca go to a lower eergy!", which it does, if it has time, ad emits eergy equivalet to the badga E g. That souds retty simle, but there are a few roblems to overcome, which all fall uder the category of iteral quatum efficiecy. There are some rocesses which are ot o light emittig rocesses. Maybe there's a "shallow imurity state", (i.e. ot at E v), the the electro gets recombied, but less eergy comes out, ad ot i the wavelegth of iterest. Or there are dee recombiatio ceters, which will also cause a oradiative recombiatio. We ca icrease our chaces of gettig radiative efficiecy by icreasig doig geerally, but oly u to a oit. The roblem is that itroducig more doats also icreases defects, which icrease the other recombiatio factors as well. Now, oce you've made light, you eed to get it out. These roblems are uder the area of exteral quatum efficiecy. Some of the light may get reabsorbed, for examle. So what to do? Oe trick is to make a heterostructure. We have a material right ext to our mai hoto geeratig material, but make it differet! If it has a larger E g, the the hotos iitially geerated wo't get absorbed by the secodary material. Oe commo material used is doed. Ad also secodly, total iteral reflectio due to refractive idex mismatch ca be a roblem. I GaAs, the small coe of from which light ca escae to air is oly about Oe solutio to this roblem has bee to isert a secodary material (such as some lastic) with a itermediate idex of refractio, ad make that outer surface sherical. This greatly ehaces the amout of light that ca escae. Laser Diodes Now for lasers, just gettig light out is't eough, you eed some serious doig to get the oulatio iversio that you eed. You eed both sides to be degeerate so that the Fermi level is withi the E c ad E v. Basically, you just wat to make sure all carriers are i the same eergy state, so that all your hoto geeratig trasitios are cosistet. Side emittig lasers E f E c E v GaAs Figure 3: Eergy bad diagram for laser diode 2002/02/15 10:21:27 Page 3 of 7

4 Figure 3 shows a eergy bad diagram for a simle edgeemittig laser. I the simlest case it's a i diode (where i is for itrisic), a it's a imrovemet o a regular diode, as a eergy well is desirable for traig the carriers so that they have a chace to radiatively recombie. This heterostructure is the defacto stadard used as the basis for desigig all edge emittig diode lasers. Figure 4 shows a simle diagram of how this laser is actually imlemeted. GaAs Light Figure 4: Simle edge emittig diode laser Usually the FabryPerot laser cavity has faces (100) so that the two "mirrors" are icely arallel. I the simle case, just the iterface betwee GaAs ad air ca rovide the reflectivity eeded. The other sides which do't articiate i the lasig ca have rougheed sides, or some other scheme to revet hoto emissio. Sice we actually get leakage out the "back" face of the laser as well, it is commoly used as art of a simle moitor circuit to verify laser oeratio (e.g. backed by a hotodiode or some other scheme). Now i ractice, there are may tricks to get efficiet lasig ad otimize various other desirable arameters. These are usually trade secrets... Vertical Cavity Surface Emittig Lasers Now for Vertical Cavity Surface Emittig Lasers (VCSEL), thigs get trickier. The idea is that the laser cavity is o loger i the horizotal lae, but is istead, vertical. The reasos that this is advatageous are due to the may issues related to fabricatio of semicoductor devices. At the curret time, fabricatio is accomlished by layig differet layers of material o to of aother, ad this is tyically doe o very large "wafers". For examle, 100's of comuter chis are fabricated at oce o 300mm wafers, which are the diced u ito idividual square die after the fabricatio is comlete. Havig active structures o the side of a device makes it difficult to fabricate, esecially i volume. It is ot ossible at this time to simly rotate the side emittig laser structure because we are caable of rovidig excellet thickess ad laarity cotrol i the vertical directio, but ot i the horizotal directio. I the vertical directio, very uiform layers ca be grow, which ca subsequetly be olished (e.g. olishig via wet slurry) to a ear erfect flatess, ad to almost ay thickess desired. I the horizotal directio, very thi structures are extremely difficult to maufacture accurately due to material fillig roblems ad the resolutio of the structures ossible whe usig light to exose hotoresist. So, the idea behid VCSEL's is fairly simle, just build u a vertical laser. That's great, but cosider the edgeemittig laser ad how it is fabricated. Suose the laser cavity is ow vertical. How does oe iject carriers ito the active regio, ad oce there, how do you have a trasaret cavity to extract the hotos? 2002/02/15 10:21:27 Page 4 of 7

5 I ractice, the actual fabricatio is very comlex. A tyical VCSEL uses may layers, ot oly to build u the active regios required, but also to fabricate the distributed Bragg reflector (DBR) structures that are used as the les ad mirrors. Oe desig might go somethig like this startig from the to layer: Coat with olyimide (mechaic stregth ad UV rotectio), somethig like C 22H 10O 4N 2. Au/Z/Au electrode tye Al 0.7Ga 0.3As/GaAs DBR (21 layers) AlAs oxide aerture for otical cofiemet (80m) The active regio cosistig of 3 layers of alteratig GaIAs (8m) / GaAs (10m) to defie 3 quatum wells (GaAs used as the barrier) AlAs oxide aerture for otical cofiemet (80m) tye Al 0.7Ga 0.3As/GaAs DBR (21 layers) GaAs substrate Bottom AuGe/Au electrode Quatum Wire/Quatum Dot Lasers Figure 5: Samle VCSEL Why do we wat quatum lasers? Well cosider this, i order for lasig to occur, we eed a certai miimum curret. This curret ca be described by: where: q = electro charge N ij = ijected carrier desity d = active layer thickess r = carrier lifetime W = laser cavity width L = laser cavity legth [Equatio 4] Lasers are't terribly efficiet, due to oradiative recombiatio ad various other losses. By isectio of the above equatio, oe ca see that to reduce the curret required (ad hece, icrease efficiecy), oe strategy is to reduce the cavity size (either width, legth or layer thickess, or all three!). That's why quatum lasers are of iterest. 2002/02/15 10:21:27 Page 5 of 7

6 Laser Diodes i NMR Sectroscoy Why? Because we ca use a olarized laser light to iduce more si i a material tha would otherwise be ossible with just magetic fields. At the curret time, i order to use this effect, we have to setu a exteral laser source. Not oly is this exesive, but it is't as robust ad is more troublesome to setu ad has some techical roblems whe oe wats to erform measuremets at very low temeratures. By usig a diode laser, we could fabricate a array of laser diodes ad have that directly i cotact with the samle we wat to measure. It is otetially much cheaer ad more coveiet. So, what do we wat i a laser diode? Well, for starters, we eed to have a stable, olarized source of laser light. Souds easy, ad it should be, but this has serious imlicatios for the choice of laser diodes. Side emissio tye lasers, by their hysical cofiguratio, icely favour oe mode of olarizatio that is redictable ad stable. The disadvatages are the oes listed for VCSEL vs. side emitters, i terms of efficiecy, form factor ad maufacturig. O the other had, while VCSEL's exhibit may desirable roerties, amog them a icely circular, low diverget beam atter. However, the same symmetry gives us a roblem, i which the olarizatio is ot tyically redictable, ad may ot eve be stable. The reasos behid this are ot well uderstood at the curret time. There is some research to idicate that articular circuit geometries or eve the switch from a (100) GaAs substrate to (311) GaAs substrate for MBE grow VCSEL's ca stabilize the olarizatio. Quatum well/quatum dot lasers are really cool, but the advatages they offer do't cocer our articular alicatio, where all we wat is icely olarized coheret light. Certaily they are very troublesome to desig ad fabricate at the curret time. However, we should ote that oe of the methods that have bee attemted i the ast at makig quatum wire lasers ivolve usig a very strog magetic field to cofie carriers o regular diode lasers. This is obviously ot feasible for mass roductio tye desigs, ad it seems very little work has goe ito this. If we are to utilize regular sideemissio laser diodes, we will most likely have to do some sort of miimal characterizatio of the laser erformace. If we have the data, it might be worthwhile to ublish some results as well, as it is't tyical for most researchers workig o laser diodes to have access to such magetic fields. Previous results have idicated ossible reduced temerature sesitivity of the threshold curret. The same lack of requiremets ca also work i our advatage. I fact, we do't eve ecessarily eed to extract ay hotos because we do't wat the light. We oly wat it to iduce greater si olarizatio, ad if this ca be trasferred from the hotos i the active regio to some of the bulk material which makes u the laser ad ito the biological samle, that's all we require. Much of the comlexity (for examle i the DBR) is required to get roer lasig ad to efficietly extract the hotos. 2002/02/15 10:21:27 Page 6 of 7

7 Glossary Term CMP CVD DBR LED MBE NMR VCSEL Defiitio Chemical Mechaical Polishig Chemical Vaour Deositio Distributed Bragg Reflector Light Emittig Diode Molecular Beam Eitaxy Nuclear Magetic Resoace Vertical Cavity Surface Emittig Laser Refereces Semicoductor Lasers I Fudametals, Eli Kao, 1999 VerticalCavity SurfaceEmittig Lasers V SPIE Proceedigs 2001, Ket D. Choquette, Chu Lei Priciles of Electroic Devices, Bart J. va Zeghbroeck, 1996 Itroductio to Microelectroic Devices, David L. Pulfrey & N. Garry Tarr, 1989 Microelectroic Devices, Edward S. Yag, /02/15 10:21:27 Page 7 of 7