IEEE TRANSACTIONS ON ELEOTRON DEVICES, VOL. Y>D-21, NO. 12, DECEMBER 1974 ' 785

TheSemiconductor Field-Emission Photocathode

DIETER K. SCHRODER, MEMBER, IEEE, R. NOELTHOMAS, JAMES VINE, AND H. c. NATHANSON, MEMBER, IEEE

Absfracf-The recently developed large-area field-emission photo- dense, needle-like emitters on flat silicon substrates [S]. is described. It consists of a finely spaced array of point We haverecently succeeded infabricating large-area emitters fabricated by etching ofp-type silicon or other semicon- arrays (up to 6-7 cm2) of p-type silicon field emitters ductor. Uniform emission over areas of 6-7 cm2 have been obtained. [SI. For Si, the spectral response extends from 0.4 to 1.1 pm. Quantum They are photosensitive and emit into yields of 25 percentat 0.86 pm havebeen measured, which is with an except'ionally high degree of uniformity. Broad- about five times the value reportedfor the extended S-20 photo- band photoemission has been observed with responsivities cathode and comparable to the best 111-V photoemitters. Calcula- at visible andnear infraredwavelengths comparable tions indicate that quantum yields of up to 40 percent at 0.86 rm and to those of the 111-V negative--affinity photo- 28 percent at 0.9 pm are attainable with the present photocathode structures. Forlow dark current densities, photocathode cooling emitters. In the transmissive mode of operation, primary to temperatures approaching77 IC must be employedat present. quantum efficiencies of 25 percent at 0.86 pm were mea- The dark current is shown to be dominated by surface-generated sured.The high photoresponseis at'tributed Do a high electrons in the space-charge region of the emitters. Effects of phos- tunnelingprobability at the surface and long minority phorus gettering and annealing treatments on dark current are dis- carrier diffusion lengths in the high-resistivity p-silicon. cussed, and the spatial frequency response of the device is determined. Theresults of a computer study showthat the field The field-emission photocathode is unique in that photo- intensification factor of p- field emitters behaves emission is observed without the use of cesium or cesium- quite differently from that of metallic emitters. oxide activat'ion,unlike other types of photosurfaces. Consequently, high-processing tenlperatures and stringent high-vacuum conditions arenot required, thus making I. INTRODUCTION the silicon-array field-emitter conlpat'ible with present-day image-tube fabrication methods. LECTRON emission intovacuum from semi- In this paper, theprinciple and operat'ing characteristics E conductling field emitters under the influence of an of the field-emitter photocathode a,re given. In Section external has been extensivelyinvestigated 11, the physics of photosensitive field emission from during the past few years. These studies were performed p-typesemiconductors is described. The field-intensi- using single-field emitters and indicate clearly that field fication factor, /?,is analyzed in Section I11 by computer emission fromp-type Si [l], Ge [a], andother semi- studies andit is shown that it behaves very differently from conductors [3],[4] is stronglyphotosensitive. Photo- that of metallicemitters. /? is high when the p-emitter electric yields approaching and even exceeding unity have operates in the Fowler-Nordheim region and drops when beenreported for Si [S] and Gc [2],[5]. While single the operating point shifts into the saturation regime, thus emittersare useful for investigativestudies, they are establishinga self-ballasting feedbackmechanism. The hardly practica,l as field-emission photocathodes. However, dark current is discussed in Section IV and is shown to be an array of field emitters such that the electron current the result, of thermalgeneration of electron-hole pa,irs from each emitting tip is proportional to the light falling in the depleted space-charge region a,t the emitting sur- onit,, presents a novel approach to photoemission, and face, domina,ted by surfacegeneration. The effects of such a device could be of considerable importance since phosphorusgettering and annealing treatments showing phot,oemission using the field-emission effect does not the importance of are treated in Section V. havea long wavelengththreshold as encountered in In Section VI, the spectral response and quantum yield cesiated 111-V photocathodes [6]. are given, both experimen.tally and t'heoretically.It) is Attempt's have been made in the past todevelop photo shown thatthe field emitterphotocathode is the only field-emission arrays. Ribik et al. [7] ha,ve reported existing cathode with eflicient photoresponse beyond 1.1- efficient photoemission from small-area (0.5 cm2) field pm wavelength. Finally, in Section VII, the spatial fre- emittersmade of Si a,nd Ge, andArthur and Wagner quency response is analyzed by considering both lateral employed the VLS epitaxialgrowth technique to form minority carrier diffusion and the discrete nature of the device. It is shown t'hat modulationtransfer fun.ction &Innuscript received ?;lay 17, 1974; revised August 8, 1974. This of 40 percentat 15-20 fine pairs/mmare possible work was sttpported in part by the Advanced Research Projects A~~~,~~and mo,litore(~by the ~Tjsion Laboratory, Fort with existing devices. For opt'imized designs, 65 percent at Belvoir, Va.,under Contract DAAK 02-72-C-0399. 20 line pairs/mm and limiting resolutions of SO line pairs/ The authors are with the Westinghouse Research Laboratories, Pittsburgh, Pa. 15235. predicted. are mm

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. 786 IEEE TRAXSACTIONSON ICLECTRON DEVICES,DECEMBER 1974

11. FIELD EMISSION FR.01LI p-TYPE t Field emission from metallic emitters is well described bythe Fowler-Nordheim (F-N)theory [lo] andthe emission current is given by

fl'2 I, = 1.54 X 10-6-AA, exp (-6.83 x lo7~$~'~v(y)/E) 4 (1) where E is the electric field, C$ is the , A, Inverse Voltage is the emitt'er area, and v(y) is a slowly varying funct,ion Fig. 1. Current,-voltage charactjerktics of a p-typesemiconductor [11] that represents the Schot,tky lowering of the poten- field emitter showing the effects of light and temperature. The tial barrier. The preexponential term is the supply func- band diagrams correspond to regions I, 11, and 111, tionand the exponential term is the transmission co- efficient. For C#J N 4-5 eV, fieldemission is normally Almost all experiments in the past havebeen performed observed at E N 1-5 X lo7 V/CIYL Plots of log (I,/E2) withindividual emitters, both for metals and semi- versus l/E are linear. It' isusually more convenient to conductors. When dealing with the simultaneous emission. plot log I, versus l/Ti~,which is also linear to a good ap- from a multiplicityof emitters where va,riationsin emission proximation. In addition to metals, n-type semiconductors fromindividual emitters may exist, the questionarises generally also yield linear F-N behavior [12], indicating as tjo whether F-N behavior (as in region I in the case that asfor metallic emitters,the supply of electrons within of p-Si) is t,o be expected at all. This problem has been the emitter is sufficient for the emission to be determined treated byTornaschke and Alpert [l6], who, by numerical only by the transparencyof the surface barrier. methods, were able to show that groups of up to 100 Field emission fromp-semiconductors is considera,bly emitterswith randomly selected field enhancement more complex for several reasons. Emission can be stronglyfactors and emitting areas do indeed yield linear E?--N influenced by the state of the surface, field penetration plots. Ample experiment'al confirmation of this result, into the semiconductor, Limit,ed a,vailability of electrons, is provided by our data [17) which indicate conclusively and by the fact tha,t emission can arise from botJh con- thatthe simultaneous emission fromalmost a million duction and valence bands. Deviations from a F-N de- emittershas alinear F-N dependenceon volta.ge. It pendence are usuallyobserved for p-typeemitters [11. would thereforeappea,r that, analyses of individual Theoreticaltreatments of fieldemission from semicon- emitters are appropriate to arrays of many such emitters. ductors have been attempted [la] and numerical solu- tionsdo indeed depict cert'ain deviations from F-N 111. FIELDINTENSIFICATION behavior, butin many important aspects,experimental The electric field, E, appearing in the F--N expression observations are not adequately described by theoretical (1) is that at the tipof the field emitter. E can be written predictions. as The generad features of field emission from p-type semi- E == fjEV (2) conduct'orshave been discussed by others [13],[14:] with andare summarized inFig. 1. The four important' re- EV = .Vv/d (3) gimes of operation are shown by regions I-IV. In region I, there is a sufficient number of electrons in the conduc- where d is the anode-to-emitter spacing, VVis the voltage tion band, and theemission current is determined only by appliedbetween anode and emitter, and Ev is tJhe field the tunneling probability at the surface. For the higher due to Vv if the emitt'er surface were planar. It has gen- voltages in 11, field penet,ration creates a depletion region erally been assumed that /?, the field intensification and causes the current to be limited by t'hesupply of factor, depends on1.y on the geometryof the emitt'er andis electrons and not the transparency of the ba.rrier. At the independent of the applied voltage. This is clearly justified still highervoltages of region 111, the field penetration for metallicemitters, where the log (current)-inverse is generally considered tobe sufficiently strongtha,t voltage relationship is linear. Quite a different situation impact ionization in the space-charge region (scr) causes isfound in p-type semiconductors, where the field can the rapidcurrent increaseshown. Since in both TI and penetrate into the emitter and its surface is no longer an I11 the supply of electrons is limited, externa.1 excitation equipotential surfa.ce. and temperature will cause the current' to increase. With In order to determine the dependence of ,6 on the geom- further increases in appliedvoltage (region IV),the etry and the applied voltage, a detailed computer study rapidlyincreasing emission current will again become was carried out. The results are presented in this section restrictedby the surface-vacuumbarrier transparency for both metallic a5nd p-semiconductorfield emit'ters. It is ~51. shom-n tha,t while for ametallic emitter 6 is inversely

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. SCHRODER et al.: FIELD-EMISSION PHOTOCATHODE proportional to the tipradius, for a p-type semiconductor, it is much lower and quite insensitive to the radius, once adepleted scr develops. n-typesemiconductor emitters behavelike metals, provided the elect'ron emission is governed by the barrier transparency and not the supply of electrons within the emitter [18]. The computer study was carried out to find an exact solution to theshape of the sor, the electric field dist'ribu- la J tion in and around the emitter and fieldthe intensification factor, by using a relaxation method to solve the partial differential equ.ations. Shrting with an infinite array of identical emitters all under the same electrical conditions, each emitter may be regarded as being at the center of a rectangular cell, extendingfrom the baseplane of the silicon t'o the anode, as shown in Fig. 2 (a). A11 the cells ibJ z-axis are identical and it is onlynecessary to solve the elec- Fig. 2. (a) Basic-problem cell. (b) Reduced-problem cell used in trostatic field problem in any one of them, imposing the the computer analysis. boundary condition of zero normal potential gradient on the cell faces that adjoin neighboring cells, i.e., of complete relaxation solutions each made with a fixed av/an = 0. (4) charge distribution while the charge distribution is modi- To reduce the three-dimensionalproblem to one in fied in an outer loop. twovariables, the cellwas approximated by acircular After exploring several very simple a,pproximations to theactual emitt'er geometry,t'he approximate shapes cross section of diameter S, imposing (4) on the cylindrical surface. The problem is then one of axial symmetry in shown by t'heheavy lines in Fig. 3were used. The structure is 10 pmhigh, tapering to a flat tipof 0.33-pm radius, and coordinates r and 2, as shown in Fig. 2 (b) . the separation between emitters is 20 pm. The model was The electrostatic potential obeys Laplace's equation set up ona finite difference net with 3 meshes per pm. The v2v = 0 (5) anode position and potential were arranged for an average in the vacuum. As a result of the carrier depletion, there field strength of lo5 V/cm. The angularcontour of the exists a chargein the silicon where t'he potentialis governed emit'ter is the result of limitations of the computer pro- by Poisson's equation gram to representcurved interfaces. For the semicon- ductor, K, = 12 was used, while for th.e metallic emitter v2v = ._PIKto. (6) calculations, the potential on t,he surface of the emitter was held at zero, equivalent to no field penetretion. The charge density p varies with voltage and was taken The lighter lincs representcalculat'ed equipotentials, to be and their shape for the metal, shown in Fig. 3(a), is as expected. For the p-semiconductor of 10 Q-cm, there is P = p0[1 - exp (- S)] significa,nt field penetration, as seen in Fig. 3(b). Using the V = 0.005 V equipotent,ial as the edge of the scr, the with po = ~NA.Surface-state charges were not considered width of the scr is 1 pm in the "valley" between emitters, inthis analysis. Atthe silicon surface, the equations but considerably larger in the emitter under the tip. This appropriate to a interface were applied. is the result of the field enhancement. The equipotential With this approximation to the charge density, there is lines and scr of a met,allic and a p-semiconductor for a nosharply defined depletion region edge.However, modified emitter structure areshown in Fig. 3(e) and (d), calculations have shown that for V 2 0.1 V, the silicon respectively. Here the tip was sharpened from a flat top is practically totally depleted while for V 5 0.001 V, t'o a 45" shape, where the triangulastip falls on the there is almost no depletion. In thecomputational method boundary of one mesh square. A significant increase of fi adopted, (7) isincorporated directly into the finite from 9.6 to 18 is observed for the metallic point, while difference relaxation formulas, so that the potential value for the semiconductor, it increases from 5 to only 5.2. resulting from one relaxation cycle is used to determine Fig. 4 shows a more detailed comparison of the field the charge density at the point for the succeeding cycle. intensification factor fi for metallic and p-semiconducting Thusthe depletioncharge distribution adjustsin step field emitters ( and open circles, respectively), as a with the potential distribution as the solution proceeds, function of theemitter height to tipradius ratio, h/r. untileventually a self-consistent solution of (6) is ob- The emittershape of Figs. 3(c)and (d) was assumed. tained to within the desired degree of accuracy.This It is apparent that for metallic emitters, increases with conceptprovides computational economy as compared increasing h/r values, as expected, and the relationship with a possible alternative method involving a sequence /3 = 1 + 0.36 h/r, was established empirically, assuming

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. IEEE TRANSACTIONS ON ELECTRON DEVICES, DECEMBER 1974

depletion conditions exist in the emitters (i.e., for opera- tion in the source-limited region 11), @ is only 5 and the electric field atthe emitter tip (5 X IO5 V/cm) would befar too low for field ernission to be observed. This stronglysuggests, then,tha.t the formation of adeeply depleted scr region within the emitter as shown in Figs. 3(b) and (d) does not occur. Instead, the semiconductor interior is shielded, presumably by tho presense of charge (fixed or mobile electrons) on the surface of the enlitter, c .SI0 and @ isthereby increased to levels which permit the X occurence of field emission. The effect of surface charge on p was investigated by computer calculations a,s follows. The surfacc charge over 5 theentire surface of a metallic-likeemitt>er was deter- mined for an h/r value of 50 and shown a.s point .A on Fig. 4. This is the minimum charge required forthe voltage 0 atthe surface to be zero. The applied fieldwas then 0 5 10 0 5 10 raisedwith the charge held constantand @ was recal- Distance from Centar of Emitter, rn IC) (dl culated. For a doubling of the field, the 0.5 data point of Fig. 3. Computer-calculatedequipotential lines for (a)metallic, Fig. 4 resulted. The assumption is that doubling the field (b) p-silicon, (c) metallic, and (d) field-emitter structures. Thl: for constant surface charge is equivalent t'o reducing the field-intensification factor is given for each case. surface charge by a factor of 2 for constant field. Similarly, the 0.75 and 0.25 data points were obtained. What these 40 pointsand extrapolated curves show, is that asurface charge does indeed increase p, and any value of @ between

fM that of the totallydepleted and themetallic case is possible. - Inan actual case, there is possibly aredistribution of L 0 charge when the field is raised and our calculations are 220.- ", intended to show the semi-quantitativebehavior of p- 02 I semiconductoremitters. Theactual change of surfa,cc M -n = 10 charge is a complicated function which depends on dark current,irradiance, surface states, and doping concen- tration. .I-L 0 40 100 to 80 In light of the @ dependence on field penetration (hence hlr on anode voltage) and surface charge,the following model Fig. 4. Calculated field-intensification factor for metal (0)and p-silicon (0)in depletion and for varioussurface charges (8). for the operation of ap-semiconductor field emitteris For metallic emittersthe theoreticalline p = 1 $- 0.36 h/r is proposed. In region I of Fig. 1, wherc p-silicon emitters also shown. are well-described by the F-N tunneling theory for metals, thereare sufficient electrons for emission to bebarrier that the emitters have a semicircular geometry as shown limited.Thus the surfacecharge of mobile electrons is in the insetin Fig. 4. However, in the case of fully depleted high and so is the field intensification. When the supply p-type silicon 2mitters, @ is generally very low (-5) of electrons decreases (limited by thermal generatio11 in and independent of the hll. ratio of the emitters. Thus region 11), the surface cha.rge diminishes and field pene- in field emission fromp-semiconductors, p depends not tration into the semiconductor takes place. This in turn solely onthe geometry of theemitter and the emitter causes p to drop. spacing, but is profoundly influenced by the penetration The p lowering acts as a feedback mechanism. Electrons of the field into the semiconductor in contrast to metallic emitted from the surface-charge reservoir at the tip cause emitters where there is no field penetration. This is an more field penetration, which in turn causes @ to drop, importantresult and some of itspractical implications thereby decreasing the ba,rriertransparency and the will now be discussed. tunneling probability. The electronemission decreases and Experirnentally, the h/r ratio of the silicon emitters was a steady-state condition is established. It is quite likely determined from high resolution SEM examinations to be that surface states influence the device behavior in addition inthe range of 1000 (h = 10 pm and r 100 A). With to these effects. Some indications are discussed in Section average applied fields of IO5 V/cm, tunneling fields of 3 V ort surface states, but their behavior is not well under- to 4 X lo7 V/cm would therefore be expected atthe stood at present. emittertips provided thatthey exhibitametal-like In region 111, avalanchebreakdown creates a higher emission behavior, as it was tacitly assumed in the linear surfacecharge which leads to larger 6, higher barrier F-Nregion (region I inFig. 1). If, however,full scr transparency, and stronger emission. This cont'inues until,

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. SCHRODEH et al.: FIELD-EMISSION PHOTOCATHODE in region IV, the currentis entirely limited bythe tunnel- ing barrier. t+t++t_tt+ t t + + t tttt-kt Another important aspect is that quite uniform emission -7 LiT7 has been achieved from large-area p-silicon field emission arrays, a consequence of the fact that variations in tip radius (or emitter height) have a negligible effect on p underdepleted scr mode operation. Since, in practice, Surface effective p’s of 100 or higher are necessary for observable States field emission, it seems likely that depletionis quite ..NA Acceptors shallow and therefore, emission isprobably insensitive to variationsin emitter gaometries oversmall ranges. Q Fortunately, suchsmall-range variations are within the capabilities of the present fabrication processes. In con- trast’,very poor emission uniformityhas been reported f from various large-area metallic emitters [19]-[21]. Fig. 5. Model of the field emitter used for dark-current calculations. The p dependenceon field penetration also explains the fields in the semiconductor. In the F-N region, for between emitters, a result of the geometric field enhance- p’s of 100-500 and averageapplied fields around lo5V/cm, ment a.s discussed in Section 111. the tip fields are in the1-5 X IO7 V/cm range. The field in The dark current is considered to be due to thermally the silicon is 1/12 this value and lies in the1-4 x 106 V/cm generated electron-hole (e-h) pairs in the scr, both in the range. Such high fields can exist in a semiconductor only bulk and at thesurface. Since there can be a scr along the if there isa sufficient density of charge at the surface. entire device, i.e.,in theemitters and also between For example, it is well known that electric fields in the emitters, the volume in which bulkgeneration takes seniconduct#orof metal-oxide-semiconductor devices reach place is given by the width of the scr and the area of the values inthe mid-106 V/cm rangeunder sufficiently entire array. Similarly, the area that determines surface strong inversion or accumulation conditions. In the case generation is considered to be the entire area of the device, of a field emitter operating in the F-N regime, there is a shown schematically in, Fig. 5. sufficient surfaceconcentration of mobile electrons to As shown in (2) and (3), the field at the emitter tip in terminate these high fields. If p were independent of the vacuo is given by anode voltage, then as the currentbecomes limited by the supply function (scr formation), isit reasonable to assume that the surface electron concentration would tunnel into where p isvoltage dependent. The field in the semi- vacuum and the high fields would have to be absorbed conduct’or, Es, is diminished by the dielectric constant, in the depleted scr. However, fields in the mid-106 V/cm Ks, and is also affectedby any surface state charge rangecannot exist ina depleted scr without avalanche QSS, as breakdown. Our model thus suggests that as the surface charge decreases (deviationfrom E’--N operat’ion), p .Es L- E/K-s =b QSS/KSE~ (9) dropsand the semiconduct>or surface field decreases where minus sign applies to negatively charged surface proportionately. In fact, p and surface charge adjust to states while, for positivesurface-state charge, the plus a steady-state condition in region I1 which is sufficient sign holds. ES is related to the semiconductor bulk charge for emission butnot for breakdown. This self-limiting by Gauss’s Law adjustment between p and surfacecharge cannot. be explained if is not allowed t’o vary. TJsing (8), (9), and (10) in. addition to the fact that the IV. DARKCURRENT anode voltage, VA,is the sum of the vacuum and semi- In the operation of the photosensitive field emitter, a conductor voltage drops, the scr width becomes positive voltage is appliedto the anode as shown in Fig. 5. This gives rise to various electric field strengt,h com- ponents: E’, arethose field lines terminating on mobile where Nss is the density of surface states per unit area. electrons at the surface of the emitting tips, E2 lines ter- The significance of the minus sign is tlhat for negatively minat,e on negativelycharged surface states,and E, charged surfacestates, the field lines terminatepartly terminate on ionized acceptors. To satisfy Gauss’s Law, on surface states, as depicted in Fig. 5. When the second it is obvious that unless there is a suffiriently high surface- term in (11) is equal to or larger than the first, it merely state charge to shield the interior of the device entirely, indicates that W = 0, since none of t’hefield lines penetrate there must be a depleted space-charge region of width W. into the semiconductor and no space-charge region exists. The scr width varies as shown in Fig. 5, because the field The scrwidth W of (11) is thewidth of the space- at theemitters is significantly higher than in the“valleys” charge region inthe planar region between emitters.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. IICEE TPIANSACTIOKS ON &LECTRON DEVICES, DECEMBER 1974

At the emitter itself, field intensification gives a higher field and W can be approximated by

w = toflVA/qNAd & -!?ss/N~. (12)

The dark currentJd is considered to be theresult of therm,tl generation in the space-charge rcgion bulkand surfa,c:e and is given by E221

Jd = qniW/r, + qniso (E) where rg and so are the generation lifetime and surface- generationvelocity, respectively [24]. Subst'itution of (12) in (13) gives

Jd = nicoflVA/r,Nad rrt qniNss/r,ilTA + qniso. (14) A physical interpretation of (14) is as follows. Tke first term representscurrent generat'ed inthe spacc- charge region, which exists in theabsence of surface state 3. The second term accountsfor changes in W brought abollt by surface states. Clearly, the sum of thesetwo term must always be positive or zero. The third termis surface generation and applies as long as the surface is depletetl. It should be pointed out that the current expression (14) is a first-order calculation. It was calculated for the plane-parallel case and then modified by the field-inter.- Fig. 6. The effect of temperature on thedark current of a 10 a-cm, p-type emitter array, unoxidized. A = 0.5 cm2, d = 3 X sification factor of the emitters.Computer calculation:r, cm. previously discussed, have shown thatthe field inter.- sification factor depends on voltage when a scr exists in theemitter. The value of ,B adjusts itself to keep the r, OK field at the tipapproximately constant. 300 280 260 240 220 2W 190 The dsrkcurrent considered previously is only thermally 10-6 generatedcurrent. We have neglected the effect of the tunnelingprobability of theseelectrons. This is a gool assumption, for ourexperiments have shown that elec- trons optically generated within the emitter array have % near unity probability of being emitted. We assume the 10-8 same is true of thermallygenerated electrons. Another factor neglected in (14) is the electric field shielding b,v. w mobile electrons at the emitter, about which very littl:: 0. 5 10'~ is known. i The most obvious test for. the dark-current theory is its dependence on temperature.In (14), ni is the most temperature-sensitiveparameter. Emitterarrays wer 3 fabricatedas described in [as]. For thedark-current' measurements, areas of 0.5-1 cm2 were etched, while for 10'11 responsivity and modulation-transfer-functionmeasure- ments,large area (5-7 cm2)arrays were fabricated. Experimental dark current-volta,ge curves are shown in 10-12 Fig.6. By plotting the current)in the source-controlletl 3.4 4 5 IOMlfl, K-' region I1 as a function of inverse temperature in Fig. 7, Fig. 7. Dark current as a function of temperature for the device straight lines are observed. The activation energy of 0.55 of Fig. 6. The activation energy corresponds to EG/~. eV is approximately equal to &/2, i.e., it is the activation energy of ni,confirming that,generat'ion ofe-h pairs ils determined by generation cent'ers near midgap either in current densities of 10-14 A/cm2 should be obtainable at the bulk of the scr or at the surface. 150 K. However, suchlow dark levels may be unattainable It is noted that the lowest current densities that haw in practice, since contributions from valence-band tunnel- been measuredare restricted to t'he 10-l2 A/cm2 rangt: ing may become important at, these low levels. (at 180 I<) due to spurious cold emission from supports Thedark current, according to (14), isinversely etc.Ext'rapolation of thedata of Fig. 7 suggests thal, proportional tothe bulk-dopantconcentration, NA.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. SCHRODGR el a1.Z FIELD-EMISSlON PHOTOCATHODE 79 1

1- o Illuminated , ;,Dar; LZZ8, b,\l ,; 1oV9 108 6 4 3 25 2 1.7~10~ vA' voits Fig. 8. The effect of resistivity on thedark current of p-type emitter arrays. A N 0.5 em2, d = 3 X IOp2 cm. Fig. 9. Current as a function of voltage andlemperature for the device of Fig. 6. Curves I, 11, and I11 are discussed in the text and represent the dark-current model. I--V characteristics of emitter arrays of resistivities vary- ing between 1 and 1000 62. cm (NA = 2 X 1016 to 1.3 x cm2, the dark-current expression is lOI3 are shown inFig. 8. It is apparentthat the current decreases with increasing doping, but the inverse relationship bet'ween I and Ar~is not strictly observed. The lifetime was measured on test wafers to be 10 ps. Bulk scr generation is characterized by the generation rg For W varying between 1-30 pm (30 pmbeing the thick- lifetime, T~,in(14). We investigated the effect of 7g ness of the wafer), the bulk-currentcomponent varies on the dark current bymeasuring the I-V characteristics between 6.5 x and 2 x lo-' A at 294 K, lower than of devices before and after phosphorus gettering. This was the experimental values. Furthermore, it wa,s mentioned done using the POC13 technique [24] which is well proven earlier that gettering, which improve:s the generation forimproving the lifetimein silicon devices [25]. The lifetime and hence should lower the bulk current compo- lifetimes were measured by thepulsed MOS (metal-oxide- nent, had no effect on the dark current. This indicates semiconductor) capacitor technique [24] on test samples thatthe surfacecurrent is the dominantcomponent. that underwentidentical treatmentsas the emitter In order tomatch the experimental darkcurrent to arrays. These measurements showed that 7gcx 10-30 ps (lj),SO was chosen to be 2000 cm/s to give afterthermal oxidation, which increased to 150-200 ps following the POC13 getteringstep. However, no corre- spondingdark-current decrease could observedbe indicating that such heat treatments, which are known This value for SO is reasonable for a silicon surface that to lower the da,rkcurrent in conventional silicon p-n has not been treatedto reduce thefast surface state junction and MOS devices, do not seem to influence the density, and according to [22] corresponds to a densi-ty darkcurrent of emitterarrays. It appears,therefore, around 1013 cm--2. as if the surface current component. is the dominant one. Equation(16) is independent of voltage and merely Inthe derivation of (14), anumber of parameters indicates the thermally generated dark current, provided were used which are not well understood, nor can they the surface is depleted. It is plotted in Fig. 9 as a funct'ion be easily measured. For example, the density and type of of temperature as line I. It matchesthe experimental surface states is not known; neither is the surface genera- darkcurrent only overa restricted voltage range. By tion velocity. In view of these difficulties, anduntil including an extrapolatedbreakdown current curve 11, more refined techniques are available for calculations and the sum 111 = I + 11 approaches the experiment,al data measurements of some of the parameters of (14), we use at high voltage. However, for lower voltages, there is (13) for dark-currentcslculations. Here q and ni are some deviation. A possible explanation is that,, inour known and reasonable valuesfor W,T~, and SOare available. model, the surface is assumed to be initially nondepleted For example, T~ can be measured on test wafers that are and thendepletes abruptly. Inreality, the depletion region exposed to heat treatments identical to those of the field forms more gradually, starting at the tip where the field emitter wafers. is highest andgradually deplet'ing theentire area. The The dark-current model is applied to the experimental area is thus voltage dependent, and the current gradually curves of Fig. 6, as shown in Fig. 9. For an area of 0.5 deviates from the F-N line.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. "Ci , 4 32 IEEE TRANSACTIONS ON ELECTRON DEVICES, DECEMBER 1974

In summary, thefollowing model of dark currentgener2,- inthe semiconductor, NE, depends on the field in the tion is proposed. At low anode voltages,the currentfollous semiconductor as the F--N dependence. Forincreasing voltage, the area ES = pNs/Kseo = 1.5 X lO+Ns (17) anuund thetip depletesand the currentis no longer linlit'ed by the barriertransparency, and the thermally for silicon. Except for the very tips of t)he emission array, generatedelectrons begin to dominateas the supply where the field is highly intensified, the vacuum field is fuxtion. Tho depleted ajrea increases gradually as a simply given by (3), i.e., E'V = Vv/d * 2 x lo5 V/cm, furlet,ion of voltageand thedark current exhibits % which corresponds to Es IV 1.5 X lo4 V/cm. From (17) sirnilarly slow deviat'ion frorn the P-N line. Eventual1 7 this requires Ns IV 10" CEI-~. If the surface-state density avnhche breakdown causes arapid current increase.. Nss 2 Ns, t,hen all the electrons pulled towards the sur- It appearstha,t in the sa,turation-like region, the dar c face are capt,ured by surface states. If, however, there are currentis surfaced-generated current.This isin accorcl no surface states, then thesurface electron charge required with our observations that>treatments that influem? to satisfy (17) is held as a.n inversion layer, Le., the bulkbehavior (e.g., phosphorusgettering) have n3 electrons are mobile. The two cases of a device with and cffect on the dark current, while those that affect the sur- without surface states are illustrated schematically in Fig. facebehavior (e.g., HZ annealing)have verya prc- 10. nnrmced influence, as discussed in the nextsection. The effect of this on the current-voltagebehavior is Obviously, surface states pla,y a very important'role in th3 shown in Fig. 11. A p-type arraywas fabricated and tested, opcrut,ion of these devices. Thishas recently been COD- resulting in t)heconventional curve (a) mit.h fairly uniform firmed by measurements of energy distributions of field- emission over the emitting area. A 150-Akthick SiO, was en?it,ted electrons from p-silicon [26]. These data indicat: then grown at 800°C indry oxygen giving curve (b), emission Erom surfa.ce states energetically located betweell which is very similar to (a), indicating that the presence the and the valence band. Our measurements of the thin Si02layer does not appear to alter the surfme of dark current ale0 indicate emission from surface statet, barrier significantly. Annealing the device in hydrogen at with the dominant emission from states near midgap a3 400°C for 1 h-a technique that result's in a low surface- evidenced by the ni-temperature dependence. state density Si/SiOz interface €or conventional oxidized Si devices-has a drastic effect on the current,. The most V. SURFACE STATES pronounced change is the linearity of t'he I-V curve and Surface states have two entirely different effects on ths: t,he fact that it has become insensitive to light. A further operation of the device. They act as generation sites of experimentalobservation is a changein the emission e--h pairs when the surface is depleted. This effect is well pattern which ismuch less continuous and shows very known for p-n junctionsand MOS devices [22]. In spotty emission. After removing the oxide, the I-V curve addition, they can either enhance or reduce the electril: returns to its original shape and the emission picture be- field in the silicon. This comes about as follows. Conside. comes more continuous. We propose the following mech- surface states that are neutral when occupied by holes. anism. For curves (a), (b) , and (d) , there are surface 'These states may have discret'e levels in the energy gap o : states at thesilicon surface. Electrons are captured andno be continuouslydistributed. When theexternal field is inversion layer exists [Fig. lO(a) J. The dark current is due applied tu the field emitter, electrons are drawn to thc to thermal bulk and surface generation in the scr, and the surface where some are captured by the surface states, electrons are emitted frorn the tips nearest where they are previously occupied by holes in p-type material, making; genemted.After H, annealing, the surface-statedensity them negatively charged. This allows thern to act as field is greatly reduced and an inversion layer is formed. The terminating charges (shown as El in Fig. 5 shielding; entire array is now coupled via this layer and emission the interior of the device). t'alces place mainly from the emitterswith the highest In spite of the fact that theinternal field may below as 2, field intensification.This is possible in this case since tesult of surface-state shielding, the external field at thc electrons can be supplied to these sit,es via the inversion tipcan be sufficiently high for field emission tu occur layer. It is observed, in fact, that the current is higher 'l'his implies that surface states may behelpful by shielding eventhough the emission areahas diminished and is the semiconductor from high fields. However, they may more spotty.Further, if the currentin (c) comes from also influence p in a manner similar to the effect of mobile emitters w7ith higher field-intensification factors, then the electrons discussed earlier, i.e., raise ,8 and hence the field. slope of the log I--l/V curve should be less than that of Which effect dominates would depend on the density and the other three curves. This is indeed the case as seen by distribution of surface states. comparing the F-N portions of the curvesin Fig. 11. Surface states have a further influence on the device Photo-insensitive I-V CUI yes were always observed behavior. Byacting as capture sites for electrons, it is whenever p-type emitter arrays were oxidized and sub- obvinus that when the external field causes electrons to sequentlyannealed inhydrogen. However, when no drift to tJhe surface, enough of thern are captured to fill deliberately grown oxide layer was present, the hydrogen the surface states. FromGauss's Law, the unitarea charge anneal had no effect on the 1-V characteristics. This is in

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. SCHRODER et Cd.: FIELD-EMISSION PHOTOCATHODE: 793

another,by preventing the formation of an inversion layer. VI.SPECTRAL RESPONSI’VITY It was stated earlier thatthe currentin the source controlled region I1 of the I-V characteristics(Fig. 1) is verydependent upon the electronpopulation inthe conduction band and consequently on any stimulus that alters that population. One way of increasing the electron concentration is the absorption of photons of energy hv 2 &, generating e-h pairs, thus making the device a useful photoemitter wit,h several unique properties. In contrast to conventional X-type photocathodes [a71 and the more recentnegative-electron affinity [28] devices-both of (bl which require cesiation-the field emitterphotocathode Fig. 10. Field-emitterschematic (a) withsurface statescapturing electrons and (b) withoutsurface states allowing an inversion has no inherent long wavelengththreshold, does not layer to form at thesurface. require cesiat’ion (i.e., it is airstable),and furthermore does not need any high-field or high-temperature forming trea.tments. In addition, because of the moderately doped substrate(typically 10 Q-cm),the electron diffusion lengt’h is quite long. For reconlbination lifetimes of 10 ps t’he diffusion lengthis approximately 150 pm. Hence photogenerat’ed carriers have a veryhigh internal quantum yield, since any electrons generated in thie scr or within a diffusion length from the edge of the scr have a very high probability of being emitted. In fact, internal quan- tum efficiencies approaching unity have been measured. The model for the external quantum yield calculations is shown inFig. 12. Photonsare incident on the field emitter, operatingin the transmissive mode, from the left. The reflectivities Ro,R1, and Rz correspond to the air/ pyrex, the py:.ex/silicnn, andthe silicon/vacuuminter- faces, respectively. While Ro and R1 cause some of the light to be reflected away from the silicon, Rz is beneficial in that it allows some of the longer wavelength radiation, 10 8 6 4 3 2.5 2 1.7~10~ which has been transmittedwithout absorption, to be v , volts A reflected backinto the device for additional e-h pair Fig. 1.1. Effect of oxide and hydrogen anneal on the dark current; generation. A further component, nonabslorbed radiation (a) after etch, (bj 150 A SiOz, (c) 400°C 1 for Hz, (d) oxide re- moved. 10 ~‘2.crn p-Si, A = 0.3 cm2. reflected from the aluminized phosphor screen, hasnot been considered here. The field emitter is characterized by a neutral bulk of contrast to observations by E’ursey et aZ.[l]who found thickness t and diffusion length L, a depleted scr of width that the nonlinearity of the I-V curves disappeared when W, and a (‘dead’’ layer, 6, at the pyrex/silicon surface. the emitterwas heated to 400-450°C in vacuo. Theirswas The ‘(dead” layer had to be invoked in order to obtain an irreversible effect’ at’tributedto the formation of a agreement between theoryand experiment. The pro- “conducting skin” along the emitter. perties of the (‘dead” layer are well described by its name, To verify the inversion-layer model, boron ions wefe Le., the lifetime is considered to be so short that minority implanted into arrays which had been oxidized to 150 A. carriersgenerated within it are lost and hence cannot ‘The ions were implanted at 50 keV at a dose of loL4ern+ diffuse to the tips to contribute to the photocurrent. The intothe oxide-passivated arrays.The depth of implan- origin of the ((dead” layeris most likely the bandbending tation was around 2500 .A giving a volume concentration of of the silicon surface by thepresence of the pyrex substrate, approximately 4 X lo1* boron atoms/cm3. After implan- such that a sink for minority carriers is created. A ‘(dead” tation,the devices were annealedin nitrogen at 800°C layeris also introduced when a high-low junctionis for 1 h to restore electrical activity and then in hydrogen formed to reduce the surfacerecombination velocity, at 400°C for 1 h. Thesearrays showed nonlinear I-V s, a.t thelight-admitting side of the device [25],[29]. plotstypical of p-arrays,indicating that the implanted However, devices described here do not have such junc- p+ layer appears to have decoupled the emitters from one tions and consequently s is quite high.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. 794

p’/p high-low junct>ions [25], and theoretical curves ( 1) and (2) in Fig. 13 were computed with such s values and a reasonable value of minoritycarrier diffusion length a’s is obtainable with gettering. We expect that curve (2) is an upper limit to the quantumyield that can beexpected for transparent silicon photo-field emitters at 77--90 K for a Si thickness of 30 pm. In addition to transmissivedata,, experiment>al point)s x=o I X are also shown for operation in t’he reflective mode. For Fig. 12. Model used in the spectr~l-responsivityand modulatioll- the long wa,velengths, the twosets of data aresimilar, transfer-function calculations. but for X < 0.7 pm, i.e., in the visible, the reflective yield is substantially higher, indicative of photonabsorption at theelectron-emitting side of the array. It is noted that for both transnlission and reflection quantum-yield mea- surements, the aluminized phosphor screen was replaced by a transparent NESA glass anode in order to avoid errors due to light feedback from microscopic pinholes in + aL[l - R2 exp (----2at)]- aL(1 - Rz)Dexp (-cd) the alun~inizedphosphor layer. Quantum efficiency curves were also calculated for - (1 + Rz)(n~ cosh ti + sinh ti) exp ( -at) j /D silicon devices operating at roorn temperature. These are 4- F[I - exp (--aWj-j exp (--d) (13) shown in Fig. 14 for the same device parameters as were used in E’ig. 13. The main difference from the 77 I< curves where is that a.s aresult of high absorpt’ion coeffkients, the quantum efficiency is lowered for short wavelength and increased for long wavelengths. This points out the im- and portance of a, thin “dead” la,yer. An experimental point of 7.5 percent at 1.06 pm is shown, which is the ~ptinlum that canbe expected for a device 30 prn t’hick.Curve (2) was calculatedwith bulk and surfaceparameters routinely achievable andit shows that efficiencies of (19) 70 percent at 0.9 pnl, 7.5percent at 1.06 pm, and 0.8 Equation (18) differs from that in [X)] in that absorption percent at 1.1 pm are possible at room temperature. within the depleted scrof width W is taken intoaccount ~y To complement the silicon quarltum yields, we have the last term in the equation, In the calculation, W is an calculated similar curves for germaniumphotosensitive average of the values applicable in the emitters and the field emitters, since it has been established t’hat emission “valleys” between emitters and the value of 5 prn \I ds from gerrrlanium is feasible [2],[5]. Thesecurves are used. shown in E’ig. 1sJwhere the curves, ( 1), (2), and (3) were Calculated yield curves are conqmrcd with experimen tal derived using the surface and bulk recombinatlion param- data in Fig. 13. The best fit with t,he transmissive-mcde et,ers and “dead” layer thicknesses corresponding to those data areobtained with s = 2 X lo4crn/s, 6 = 0.8 ~KWI,of Fig. 13. The optical absorption coefficients were those and I, = 50 pm. The absorption coefficients used in these of 1‘311 for 77 K. These curves show tha,t for reasonable calculationsare .those of Ihsh and Kewnnan [31] at values of 8, L, and 6, efficiencies of 10-50 percent shouldbe 77 K, since t2hedat,a at90 E(: [321, the temperature of mea- achievable to X = 1.45 pm and imaging is possible to surement, are given only for wavelengths near l prn. ?‘he about 1.6 pm. The rather abrupt drop in q at 1.41 prn is quantum-yield expression (18) represents the electrons due to theabrupt increase of CY a.t that wavelength, generated wilhin .the device. The fact that there is quite corresponding to excitation of electrons from the valence good agreement between experirne~ltand theory indica tt:s band to the conduction band by direct transition. that the escape probability of these electrons is very close Curve (2) in Fig. 13 shows that maximum yields of 40 to unity. percent at 0.86 pnl and 28 percent at 0.9 pm are possible The devices were fabricated by electrostatically bond.ng with the present silicon array st’ructure when the device CY31 p-typesubstrates to pyrex faceplates. No shallow is operated near 77 K. The 1.06-pm yield at this bempera- p+ la8yer was formed on these devices necessitating .;he ture is around 0.2 percent’,but values as high as 7.5 percent high values of surface recombination velocity in the (al- have beer] measured at room temperature. The currently culations. The “dead” la,yer is possibly the result of bmd at-tainable 1.06-pm quantum yields from thick (reflective) bending introduced by the bonding process, which causes andthin (serni-transpasent) silicon arraystructures are minority carriers to be collected at the Si/pyrex inter.f%Loe summarized 011 the absorptiorl vs. target thickness plots with a correspondingly high s value. Surface recombina- shown in Fig. 16. It is evident that the 1.06-gm absorption tion velocities of around 100 cm/s can be achieved ~ithwithin a given layerthickness decrea,ses rapidlyas the

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. Wavelength, vm Fig. 15, Theoretical quantum-efficiency curves for germa~nium field-emitter arrays. The parameters used in %he calculations of the three cnrves correspond to those of ITig. 3.3.

L- 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 Wavelength, pm Fig. 14. Theoretical quantum efficiency curves for silicon field- emitter arrays operated at 300 K. The parameters used in the calculations of the three curves correspond to those of Fig. 13. temperature is reduced. This imposes some ].imitations to Target Thickness.pm the I.06-pm performance of the semi-tra,nsparent stru.c-. Fig. 16. Plots of the 1.06-pm absorption vs. thickness for silicon at va,rious temperatures. Experimental dataare 1.06-pm qua,ntum tures which must, at present, be cooled to 90 K for low yields of silicon field-emission arrays measured in reflection, n, dark-currentoperation. The cause of the low yield at and transmission, 0. Data were obt,ained at Vanode of 4 kV except low temperatures is the increasing bandgap as the tempera- where noted otherwise. ture is decrea,sed wit!ll a, corresponding smaller absorption coefficient. reported for Ge [SI. It should benoted, however, that, The experimental quantum yield pointson Fig. 16 while such multiplicat,ive gains ca,n he usefully employed agree fairly well with absorpt,ion curveswith the exception to reduce the number of secondary gain sta,ges in an in- of the 296 K “thicktarget” data point. The latteris plotted tensifier device, the excess noise introduced by this multi- for the real target’ thickness of 250 pm. However, if tJhe plicat,ion process mag deQerioratc the operation of the rnjnority ca,rrier diffusion length is less than t,his, then photocat,hodeunder low light level imagingconditions. the effective thickness, which is just the diffusion length, The emission uniforinity, which might be expect,ed to should be used. On the 296 K, a = 16 ern-’ line, the deterioratewhen the device is operatedin th.e multi- experimentalvalue of 7 = 15 percentcorresponds to a plication mode, actua,lly improves in this modeof operation. thickness of 100 bm, a reasonable value for the diffusion Since phosphor saturalion effects arc &sent; at’ the very length of the device which was not gettered. low screenloading levels a.pplicable here, thisapparent The “multiplication-mode” data correspond to yield self limiting behaviorof field-emission arrays at high. anode measurementsperformed with the device biased at the voltages is believed to be the result of field intensification onset of the avalanche breakdown region III in Fig. 1. factor ad,just>mentas discussed previously. Substantially higher photoelectric yields can be obtained in this mode of operation, as shown in Fig. 17. Curves (1) VII.SPATIAL FREQUENCY II,ESPONSF, and (2) areexperimental and (3) corresponds to curve (1) Inthe derivation of the quantum-yield expression, in Fig. 33. In fact, efficiencies atbove 100 percenthave the incidentradiation was laterallyuniform. For the beenmeasured by us forSi andhave been previously spatial frequency responseandysis, the incident light was

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. IUEE TRANSACTJOXS ON ELECTRON DEVICIGS. DILCIOUBER1974

1 --l---T--T--- F VA = 6000 Volts 1 0.8

0.6

IL 5 0.4

0.2 Ill Transmlsrive Yleld lo-31 121 Reflective Yield (31 FractionalAbsorption In Target 0 0 4 8 12 16 Spatial Frequency, line pairdmm 10-4 Fig. 18. Calculated MTF curves as a funct,ion of spa.t,inlfrequency 0.4 0.5 0.6 0.1 0.8 0.9 1.0 1.1 for the model of Fig. 12, considering lateral elect,ron diffusion and Wavelength, Pm the discrete nature of the emitter. Fig. 17. Spectral response of point-array photoemitter (silicm (lll), 10 Q-cm, p-type, 30 ,urn thick) at YO K and VA = 6 1.V (multiplication-model).Measurements made intmnsmissi In that the wavelength can ha8ve a substantial influen.ce on (curve 3) and reflection (curve 2). The calculated fractioral absorption in target thickness is shown in curve 3. the MTF. Th.e reamn for tbis is that the shorter - lengths are absorbed closer to the surface with a corre- spondingly longer distance to diffuse through. The dotted assumed to bestationary, monochromatic, and varyillg curves are for a wavelength. of 0.9 pm corresponding to a = in intensity only in the transverse direction. 110 while the solid lines are for X = 0.55 pm with The spat,ialfrequency response is conveniently repre- CY = 3.7 x lo3 cm-'. The curves were calculated for I; = sented bythe modulationtransfer function (MTF), 50 pm and s = 2 X lo4 cm/s, the values that were used which is the ratio of the output of the device when t'l~e in thespectral response calculations. It is interesting input is sinusoidally modulated, to the output when t'lle to note that the "dead" layer thickness has essentially no input has zero spatial frequency. It is given by [33] effect on the MTF. One further observation is thak the MTFi = ~]k/rO (20) values of L = 150 prn, s = 100 CIY)/S, and t = 80 pm, where vOis given by (18) and vk is similar, except that' tlle which were used as an upper limit in the efficien.cy cal- diffusion length L is replaced by aneffective value L' [33]: culations, gave a curve which coincided with the t = 50 pm curve of Fig. IS for X = 0.55 p.m. This indicates tha.t, L' = L/ (1 + Ic2L2) 1'2. (2..) while an improved diffusion lel~gthand surface recombina- The quantity IC is the radial frequency per unit distallmx tion. velocity is required for high quantum ctficiencks, the given by IC = 27rN, with LV being the spatial frequency in. spatialfrequency response suffers hecause the carriers line pairsper unit distance.For X = 0, i.e., uniform can diffuse over larger distances. irradiation, the MTFi value is unity. The MTF curves were calculated for emitter spacings TheMTFi is the modulationtransfer function th,tt of 25 pm except for t'he uppermost) one which corresponds results from lateral diffusion of minority carriers in tlle to a spacing of 12.5 pm. It shows that for a 1O-pn1 thick undepleted region of the emitter. The actual R4TF of tl~e device with 12.5-pm emitter-to-emitter spacing, the MTF device is degraded bythe discrete nature of Dhe field at 20 line pairs/mm is expected to be 65 percent for the emitter, because emission occurs only at the emitterlocz- particular L and s values shown. These dimensions are tions. This effect of discret'e structures has been treat>c:d probably near the limit for real emitter structures, and,for in great detail by Revuz[33]. The result is that the actud sucha device, the limiting spatia,l frequency is 80 line MTF can be written as pairs/mm at an MTF of 0.05. In thecorrection term sin x/x, which takes intoaccount MTF = MTFi sin (7r.V~)/ (nZ;p) (22) the discrete nature of the device, the spa,tial pha,sc between where the MTFi termis determined by the carrier rnoticm the emittersand the light pattern was not considered. within the semiconductor andthe sin x/x termis tlle Such a, phase shift can have a very significant influence result of the discrete structure.Equa,tion (22) is vald on the MTF as shown in Fig. 19. Consider a spatial fre- for N < l/p. Since sin x/x < 1 for x > 0, it is chr quencysuch that N = 1/2p,i.e., one wavelength spans that the discrete nature of the photoemitter reduces tlle two emitter spacings as shown inFig. 19(a). For zero MTF. ph.ase difference, where the peaks and valleys of the Calculated MTF curvesare shown inFig. 18. It is modulatedlight signal coincide with t,he emitters, the quite evident that the MTFdepends quite strongly on tlke sin (nNp)/(aNp) term becomes 2/71.. Flowever, a 90" undepletedwidth t. The wider this region, the more phaseshift makes the response go to zero. Physically, lateral diffusion cantake place before theminorhy this is shown in Fig. 19 (b) where the 90" shift) causes e-h carriers reach the scr. The other observation in Fig. 18 is pairs to begenerated between emitters, result'ing in

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. LICHRODER et al. : FIELD-EMISSION PHOTOCATHODE 797

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply. 798 WEE TRANSACTIONS ONDECEMBER ICLECTRON DEVICES, 1974

[3] 0. H. Hughes and P. G. Rrist>ow,“Field emission from Gal€‘,” ductors,” Sou. Solid State Phys., vol. 13, pp. 2612-2613, April Solid State Phys., (a), vol. 2, pp. 503-509, July 16, 1970. 1972. [4] I. L. Sjokol’skaya and G. p. Shcherbakov, “Nonlinearity of t, le [19] H. E. Cline, “Multineedle field e_mission from the Ni-W eutec- volt-amperecharacteristics of field emission fromCdS sinple tic,” J. Appl. Phys., vol. 41, pp. (641, Jan. 1970. ,” Sou. Solid State phys., vel. 4, pp. 31.436, July 1962. [20] H.Pfleiderer and H. Rehme, “Spike cathode for field emission,” [5] P. G. Borzyak, A. F. Yatsenko, and L. S. Miroshnichenko, Solid State Phys., vol. 11.A, pp. 153--160, May 16, 1972. “photo-field-emission from high resistance si and G~,”Ayolid [21] R.N. Thornas and B. J. Shaw,unpublished results on Cu-Cr State’Phys., vol. 14, pp. 403-411, April 1, 1966. eutectic alloy emitt,ers. [61 seefor example,R. J~.~~11 and W. E, sPicer, (13-5 compou,,d [22] D. J. Fitzgeralcl, and A. S. Grove, “Surface recombination in photocathodes:a new family of photoemitters withgreatly irn- semiconductors, Surf.Science, vol. 9, pp. 347-369, Feb. 1968. proved performance,” proc. IEEE, vel. 58, pp, 1788-18~2, [231 R. N. Thoy:s, R. A. Wickstrom, D. K. Schroder, andH. (2. Nov. Nathanson, Fabrication and some applications of large-area 1970. silicon field emission arrays,’’ Solid-State Electron., vol. 17, pp. [7] 1‘. F. Bibik, P. G. Borzyak, and A. F. Yatsenko, “Ge and Si 155-163, Feb. 1974. field electron-emission photoelements,” Ukrainian Phys. Gr~. [24] D. K. Schroder and J. Guldherg, “Interpretation of surface and VOI.13, pp. 621-622, NOV.1968. bulk effects using the pulsed MIS capacitor,” SolihState [SI R. S. Wagner and W. C. Ellis, “The vapor--solid mectl- Electron., vol. 14, pp, 1285--1297, Dec. 1971. anism of growth and its application to Si,” Trans. Mk [251 T. M. ~~~k,H, c, casey,J, v, Dalton, and M. yamin, ‘~1~- Soc. AIME, vol. 253, pp. 10,53-1064, June 1965. fluence of bulk and surface properties on image sensing silicon [Y] R. N. Thomas and H. C. Nathanson, “Photosensitive field diode arrays,” Bell Sysl. Tech. J., vol. 47, pp. 1827-1854, Nov. emission from Si point arrays,” Appl. Phys. Lett., vol. 21, pp. 1968. 384-386, Oct. 15, 1972 and “Transmissive-mode silicon field [26] B. F. ~,~~i~and T. E. Fischer,

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 6, 2009 at 10:39 from IEEE Xplore. Restrictions apply.