Secondary Electron Emission from Multi-layered TiN/Al 2 O 3 Transmission Dynodes
H.W. Chan, V. Prodanović, A.M.M.G. Theulings, C.W. Hagen, P.M. Sarro, H. v.d Graaf
PPrepared for submission to JINST
Secondary Electron Emission from Multi-layeredTiN/Al O Transmission Dynodes
H.W. Chan, a , b , V. Prodanović, a , b A.M.M.G. Theulings, c C.W. Hagen, c P.M. Sarro, b and H. v.dGraaf a a National Institute for Subatomic Physics (NIKHEF),Science Park 105, 1098 XG, Amsterdam, The Netherlands b Faculty of Electrical Engineering, Mathematics, and Computer science, Department of microelectron-ics/ECTM,Feldmannweg 17, 2628 CT, Delft, The Netherlands c Faculty of applied sciences, Department of Imaging Physics, Delft University of Technology,Lorentzweg 1. 2628 CJ, Delft, The Netherlands
E-mail: [email protected]
Abstract: The transmission secondary electron yields of multilayered Al O /TiN membranes/filmshave been determined for sub-10-keV primary electrons. These membranes will be used as transmis-sion dynodes in novel vacuum electron multipliers. A bi- and tri-layer variant has been manufacturedby means of atomic-layer deposition (ALD) of aluminum oxide and sputtering of titanium nitride.Their transmission electron yield has been measured by a collector-based method operated withina scanning electron microscope. The total transmission and reflection yields have been determinedfor both types of membranes. The results show that the tri-layer membrane, where the conductiveTiN layer is sandwiched, performs better in terms of transmission electron emission. A maximumtransmission yield of 3.1 is obtained for a 5/2.5/5 nm thick Al O /TiN/Al O film with 1.55 keV-electrons. In addition, the transmission characteristics of the bi-layer membranes have been furtherinvestigated by separating the transmitted fraction and the transmission secondary electron yield.The latter is then normalized by its maximum yield and energy to obtain a ’universal’ transmissionyield curve. For thin films with a thickness below 20 µg / cm , the transmission characteristicsdeviates from thicker films, which can explain the higher transmission secondary electron yieldobserved.Keywords: secondary electron emission; transmission dynode; photomultiplier; vacuum electronmultipliers; atomic layer deposited alumina; ultra-thin filmsArXiv ePrint: 1234.56789 Corresponding author. a r X i v : . [ phy s i c s . i n s - d e t ] A ug ontents Vacuum electron multipliers, such as photomultiplier tubes (PMTs), employ secondary electronemission (SEE) for photon detection [1]. The detection principle is the conversion of photons intophotoelectrons by the photoelectric effect and subsequent electron multiplication in vacuum. Aphotoelectron, accelerated from the photocathode towards the first dynode, will generated multiplesecondary electrons (SEs) upon impact. The SEs are then guided and accelerated towards the nextdynode. As they traverse from dynode to dynode, their number increases, until the SEs are collectedby the anode. PMTs are one of the most sensitive single-photon detectors that are still widely usedfor single-photon detection due to its high gain, low noise and large acceptance surface. Though,there are a few disadvantages to the design. First, the timing resolution of the order nanosecondsis poor in comparison with Silicon photomultiplier (SiPM) with single-photon avalanche diodes– 1 –SPADs) [2]. It is due to the large non-uniform electron paths that the SEs need to traverse withinthe dynode stack that affects the pulse width. Also, the SEs are susceptible to electrostatic andmagnetic fields, which exclude PMTs to be used near strong magnetic fields. And lastly, the dynodestack makes PMTs large, fragile and expensive.The goal of the MEMBrane project is to develop a vacuum electron multiplier that outperformsPMTs in terms of time and spatial resolution [3]. The device, the Timed Photon Counter (TiPC),has the same detection principle as a PMT, but has transmission dynodes (tynodes) as multiplicationstages instead of (reflective) dynodes. Tynodes are extremely thin membranes where an impingingprimary electron (PE) on the frontside releases multiple secondary electrons from the backside.This distinctive property allows tynodes to be closely stacked on top of each other. The distancebetween neighbouring stages, in comparison with dynodes in PMTs, is greatly reduced and theelectric field is nearly homogenous. As a result, the time resolution improves: the pulse width andthe rise time of the signal will be smaller due to the more uniform electron paths. In addition, thesusceptibility to magnetic fields is reduced due to the increased electric field strength. In terms ofspatial resolution, 2D spatial information is gained by combining the planar tynode stack with aCMOS-pixelchip (TimePix) as read-out.A tynode is an ultra-thin membrane that (1) consist of a material with a high secondaryelectron yield (SEY), (2) is mechanically strong and (3) is electrically conductive. The transmissionsecondary electron yield is defined as the ratio between the incoming PE, with an energy E , andoutgoing SEs in transmission. For TiPC, the goal is to achieve a transmission secondary electronyield of 4 or higher for primary electrons with sub-2-keV energy. As such, mechanically strong andthin membranes are required, since the range of PEs is energy depended. In-plane conductivity isrequired to replenish the emitted electrons during prolonged operation in order to avoid charge-upeffects.The SEY of a surface depends on its material properties and surface condition. In general,dielectrics have higher yields in comparison with semiconductors and metals [4, 5]. This isattributed to the wide band gap of dielectrics which benefits the transport of internal SEs. Onceinternal SEs are promoted to the conduction band, they can travel a relatively large distance with fewinteractions, which increase their overall probability to reach the surface. Surface treatment, suchas caesiation and hydrogen-termination, can lower the electron affinity, which will also increase theescape probability of internal SEs. In some cases, even negative electron affinity (NEA) can beachieved; an internal SE that reaches the surface will encounter no barrier and will be pushed intovacuum. This is beneficial to the SEY. The reflection electron yield (REY) of C(100) diamond, forinstance, increased from 3 to 60 and 132 by Cs- and H-termination respectively [6].For transmission SEE, the thickness of the membrane is an additional parameter that affectsthe transmission electron yield (TEY). The distance that a primary electron with energy E cantravel is defined as the range and is given by R = CE n , where C is a constant that is materialdependent and n a constant that depends on the energy-range of the primary electrons [7]. Theonset of transmission SEE is expected to occur when PEs are expected to penetrate the tynode. Thischaracteristic is defined as the critical energy E c for which 1% of the PEs manages to pass through: η T ( E c ) = .
01 [8]. This coefficient η T is the forward-scattered electron (FSE) coefficient or thetransmittted fraction. A second characteristic (tied to the thickness) is the maximum energy E max T at which the maximum transmission yield σ max T is achieved: σ max T ( E max T ) . Both are unique defining– 2 –eatures of a transmission yield curve correlated to the thickness of the membrane.One of the first working transmission-type photomultipliers has been built by Sternglass et al[9, 10]. The tynodes consist of porous potassium chloride (KCl) deposited on top of an aluminumfoil. The high TEY of porous materials is due to the built-up of charge inside the pores ofthe material, which results in a strong electric field where (secondary) electrons are acceleratedinternally causing an avalanche type of SE emission. The typical inter-stage operating voltage is5 keV with a maximum transmission yield σ ,max T of 8. Despite the high TEY, the required highvoltage for a multi-stage device limits its applicability. Also, the lifetime of the devices are poorand further research in the aging mechanism was needed [11]. Other alkali halides (CsI, KCl, NaFand LiF) have been measured by Llacer et al [12]. They are deposited onto an Al/Al O membraneas support, which added to the overall thickness. The highest TEY of 8 (8 keV) was measuredfor cesium iodide. The best performing alkali halide was reported by Hagino et al. on caesiumactivated CsI. They achieved a TEY of 27 (9 keV) for Al O /Al/CsI(Cs) films [13]. A secondgroup of materials that is considered are semiconductors, such as silicon and gallium arsenide, thatbenefits from negative electron affinity (NEA). A TEY of 725 (25 keV) for a 4-5 µm thick siliconfilm with NEA was by Martinelli et al [14]. More recently, various types of diamond have beenstudied as SEE materials for transmission dynodes [15–18]. The highest TEY of 5 (7 keV) wasobtained for f-NCD diamond. Although the results are promising, it is unclear whether thinnerNCD diamond can be manufactured due to the growth process of NCD diamond.After reviewing these and other papers on tynodes, it became clear that the tynode needs to beself-supported [19]. Otherwise, it is unlikely that they will perform optimal for sub-2-keV electrons.Therefore, the choice in materials is limited to materials that are mechanically strong and has a highSEY.Our group approached the problem from a micro-fabrication/engineering point of view byimplementing MEMS technology. Silicon nitride tynodes were fabricated by low-pressure chemicalvapor deposition (LPCVD) and aluminum oxide tynodes by means of atomic layer deposition (ALD)[20, 21]. Monte-Carlo simulation has shown that the optimum thickness for aluminum oxide tynodesis about 10 nm [3]. Therefore, the ultra-thin membranes, with a diameter of 10 to 30 µm, weresuspended within a supporting mesh with an array of 64-by-64 small windows [21]. A TEY of 1.57(2 .
85 keV) was measured for TiN/SiN films and a TEY of 2.6 (1 .
45 keV) for TiN/Al O films.In this paper we will present the method that was used to determine the transmission yield. Also,new samples with a tri-layer Al O /TiN/Al O films with various thicknesses will be presented andcompared to the bi-layer TiN/Al O films. Furthermore, the emission characteristics of these filmswill be discussed more thoroughly. The experimental setup is a collector-based method developedto determine the transmission yield within a scanning electron microscope (SEM). The imagingcapability of the SEM is used to locate and to direct the electron beam on the small windows.A point of concern, for any SEE measurement setup, is the built-up of charge on the surfaceof dielectrics [22]. The recommended strategy is to limit the electron dose, which can be achievedby using a pulsed electron gun [22, 23] and/or to neutralize the charge with a flood gun betweenmeasurements [24]. A different approach is to determine the SEY by measuring the surfacecharge using the Kelvin probe method [25]. However, charge-up effects were not observed onfilms/membranes on which TiN was sputtered [21].TiN was chosen as a conductive layer to provide in-plane conductivity. The added layer does– 3 –ncrease the thickness, but has a relatively low stopping power due to the low Z value of TiN. Otherconductive materials were considered, such as metals (Al, Cr), but they will most likely oxidizeduring the fabrication process. TiN is chemical inert in ambient conditions [26]. Charge-up withinthe alumina layer was not observed, i.e. the emission current is constant during exposure. Themechanism that provide (normal-to-the-plane) conductivity from the conductive layer to the chargedregion in the dielectric film can be either explained by electron-beam induced current (EBIC) [27]and/or electron tunnelling [28].Another caveat of this method is the lack of ultra-high vacuum (UHV) in our SEM, whichoperates at 1 × − mbar instead of 1 × − mbar or lower. As a result, surface contaminationwill form after prolonged surface irradiation, which might affect the SEY [29]. The contaminationrate depends on the electron dose per unit surface, which can be lowered by scanning the electronbeam over the surface. Though, a comparison between this setup and a dedicated UHV systemhave been made by measuring the reflection SEY of a SiN and an Al O film [3]. The results werein good agreement and contamination effects were not observed. However, for dedicated surfacetermination studies should be performed in a UHV system. Secondary electron emission (SEE) is described as a three-step process: generation , transport and escape of internal SEs [4, 5]. This model can be extended to thin membranes by including theexit surface of the membrane in transmission (Figure 1). The first step of the model treats primaryelectron interaction, energy transfer and secondary electron generation. A primary electron thatinteracts within a thin membrane will scatter and lose energy. Some of the energy is used togenerate internal SEs. The primary electron itself can be reflected, absorbed or transmitted by themembrane. Reflected primary electrons are designated as backscattered electrons (BSE), whiletransmitted electrons as forward-scattered electrons (FSE). They are distinguished from secondaryelectrons by their energy, which is E se >
50 eV. The second step describes the transport of internalSEs within the material. The band gap model is used to explain the difference in transport inmetal, semiconductors and dielectrics [4, 5].The wide band gap of dielectrics allows SEs that arepromoted to the conduction band to travel a relatively large distance with few interactions. Thisincrease the probability of the SEs to reach the surface. The third step models the escape of internalSEs into vacuum at the solid-vacuum boundary. Internal SEs with sufficient energy to overcomethe work function or electron affinity can escape into vacuum. Only internal SEs that are generatednear the surface have a chance to escape. The escape probability is given as an exponential decayfunction with λ the mean free path of SEs. The secondary electrons that escape from the frontsideare designated as reflection secondary electrons (RSE) and from the exit as transmission secondaryelectrons (TSE).The elementary theory of secondary electron emission predicts the (reflection) secondaryelectron yield on the assumption that the production, transport and escape mechanisms of SEscan be treated independently. The advantage of the semi-empirical formula is that it can predictthe shape of the secondary electron yield curve by normalization. There are numerous variationsdepending on the energy loss function used in the model, such as generalised power-law model,constant loss model and Bethe-model [30, 31]. The accuracy of the formula depends on the model,but for the scope of this paper the general form of the formula is sufficient.– 4 – igure 1 : Three-step model of SE generation. The three steps are treated independently in theelementary theory of SEE. The first step describe energy transfer of PEs in the film/bulk. Thesecond step models the transport of internal SEs. The last step describe the escape probability ofSEs from the material into vacuum.The production of SEs is given by n ( x , E ) = ε dEdx (1.1)where ε is the average effective SE excitation energy and dE / dx the energy loss function. Thenumber of SEs that are produced is proportional to the energy deposition at depth x . The transportand escape mechanisms of SEs is given by an exponential decay law f ( x ) = Ae − x / λ (1.2)where λ the attenuation length of the SEs (or the mean free path of the SEs) and A is the escapeprobability of a SEs at the surface of the material into vacuum, which depends on the electronaffinity. The reflection secondary electron yield δ R is then given by δ R = ∫ d n ( x , E ) f ( x ) = − A ε ∫ d dEdx e − x / λ dx (1.3)where d is the film thickness. For bulk samples d → ∞ . The transmission secondary yield δ T isthen given by δ T = ∫ d n ( x , E ) f ( x ) = − A ε ∫ d dEdx e ( x − d )/ λ dx (1.4)Secondary electrons that are generated within the escape depth λ have the same escape probabilityin both reflection and transmission for a membrane consisting of one material.The threshold energy for transmission SEE E th is correlated to the critical energy E c . Theformer is defined as the PE energy at which the first (slow) TSEs starts to emerge, while the latter is– 5 –efined as the PE energy at which 1% of the (fast) PE passes through the film. The distance that aPE with energy E can travel within a film/bulk material is defined as the range. There are a varietyof range-energy relations [32]. The accuracy of these relations depends on the material consideredand the energy of the PE. For sub-10-keV electrons and alumina as material, the range-energyrelation given by Fitting et al [8] is the most accurate and is given by R = ρ − . E . (1.5)where R is the range in nm, ρ the density given in g / cm and E the primary beam energy in keV.The range R of a PE in different materials will differ, which makes comparison of composite filmsto normal films difficult. However, the effective layer method can be applied to films with differentmaterials [33]. The contribution to the stopping power of material 2 can be replaced by material 1with an effective layer thickness given by d eff1 = (cid:18) d R (cid:19) p / p R (1.6)where p , is the transmission parameter and R , is the range in the first and second materialrespectively. The total effective film thickness is then given by d = d eff1 + d . The bi-layer and tri-layer of TiN/Al O films can be represented by a single Al O layer with total effective thickness d ,so that they can be compared to other Al O films. For low- Z materials, the transmission parametersare assumed to be approximately equal: p (cid:27) p . The conversion factor is then simply the ratiobetween the ranges: R Al O / R TiN (cid:27) .
51, i.e. the TiN layer can be replaced by an Al O layer withan effective thickness that is 1.51 times larger. The fabrication process of the ultra-thin composite membranes is similar to the fabrication processof tynodes presented in ref. [21], but the process is simplified by omitting the support mesh.Instead, a single square membrane with a width of 400 µm is released from the substrate. Thisbasic design is not intended to be used in an actual detector, but is designed with the goal tocharacterize the transmission secondary electron emission of the multi-layer membranes. In figure2, the flowcharts of the fabrication process of two types of composites are given: a TiN/Al O bi-layer and a Al O /TiN/Al O tri-layer membrane. The conductive layer is applied as a post-processin the former (A5), while it is integrated in the process flow of the latter (B5). The additionalalumina layer serves as a protection layer against hydrofluoric (HF) vapor etch.For TiN/Al O bi-layer membrane, a 4-inch p-type (5-10 Ω cm) wafer with a thickness of ( ± ) µm is used as substrate. The Si substrate is oxidized in a wet thermal environment at1000 ◦ C until 300 nm of silicon dioxide is formed. This layer will act as a stopping layer and as asacrificial layer in the process. ALD alumina is grown on top in a thermal ALD ASM F-120 reactorusing trimethyl-aluminum (TMA) and water as a precursor and reactant, respectively (figure 2 A1)at 300 nm. The thickness is varied by choosing different numbers of cycles. Plasma-EnhancedChemical Vapor Deposition (PECVD) silicon dioxide is then deposited on the front side to protect– 6 – igure 2 : (A1-A5) Flow chart of the fabrication process of the TiN/Al O Bi-layer membrane.(B1-B5) Flow chart of the Al O /TiN/Al O Tri-layer membrane.the SEE layer and on the backside as a masking layer (A2). The silicon substrate is removed byDeep-reactive Ion Etching (DRIE) (A3). After this step, the wafer is cleaved into 15-by-15-mmpieces along predefined break lines. For the final release, the silicon dioxide layers are removed inan HF vapor etch chamber (A4). As a last step, titanium nitride is sputtered as a post-process (A5).This allows the thickness of the conductive layer to be varied and optimized. The active area of thesamples is 400 µm by 400 µm and has a thickness [ d TiN / d Al O ] of 2.5 / 10 nm, 5.7 / 25 nm and 5.7/ 50 nm.For Al O /TiN/Al O tri-layer membranes, the process is the same until after the first ALDalumina deposition (A1). After this step, a small patch of Alumina and silicon dioxide is removedby plasma etching to expose the silicon substrate (B1). Titanium nitride is then sputtered onto thewafer forming a continuous layer that is in contact with the silicon substrate. Another ALD aluminalayer is used to encapsulate this layer (B2). This encapsulation is needed to protect the TiN layeragainst HF vapor in the last step. The next steps are similar to the previous process. PECVD silicondioxide is applied as protection and masking layer (B3). The silicon is removed by DRIE (B4) andthe wafer is cleaved into 15-by-15-mm dies. The membrane is released by HF vapor etching (B5).The active area of the samples is 400 µm by 400 µm and has a thickness [ d Al O / d TiN / d Al O ] of 5 /2.5 / 5 nm and 12.5 / 5.7 / 12 . The experimental setup is designed to be mounted onto the moving stage of a Scanning ElectronMicroscope (Thermo Fisher NovaNanoLab 650 Dual Beam). A teflon holder is attached to the stagein which the setup is fixed. Teflon insulates the sample holder electrically from the stage and thechamber. The SEM has an electron source that provides a continuous electron beam with energy– 7 – igure 3 : Schematic drawing of the experimental setup. The primary beam current I is measuredwithin the faraday cup. The sample holder, retarding grid and collector are electrically insulatedfrom each other with Kapton foils. Each are connected via feedhrough to Keithley 2450 sourcemeters.ranging from 0 . × − down to1 × − mbar.A schematic representation of the experimental setup is given in figure 3. It consists of 3separate electrodes: a collector, retarding grid and sample holder. They are electrically insulatedfrom each other with sheets of kapton foil. Each electrode is connected via a feedthrough to aKeithley 2450 Source meter. This allows each electrode to be biased from −
200 V to 200 V, whilemeasuring the currents simultaneously. The sample is clamped inside the copper sample holder.Silver emulsion is applied on the silicon substrate of the samples to ensure good electrical contactbetween the sample and holder. The primary beam current I is measured inside a Faraday Cupthat is drilled into the sample holder close to the sample.The primary beam current I as a function of the electron beam energy E is determinedbefore the measurement of the SEY of the samples. The beam is stable over the course of aday, so measuring the current once is sufficient. The beam is then moved towards the membranein the sample. During the measurement, the SEM is in image acquisition mode. The imagedsurface is the only part that is being irradiated, so there is no need for corrections for inducedcurrents on the surrounding ’inactive’ parts of the sample. Also, charge-up effects are mitigatedby distributing the beam over a larger surface. The continuous surface scan has a Horizontal Fieldwidth (HFW) of 366 µm and a vertical field width of 316 µm with a resolution 1024 x 884 pixels.This is approximately 0 .
116 mm over which the current is spread. The dwell time (per pixel) is1 µs. Charge-up effects can be identified with the SEM, as the contrast of the image changes on acharging surface. Additionally, if charge-up occurs, it can also be observed in the measured sample– 8 –urrent, which will change over time. For each energy E , the surface is therefore scanned for20 s before shifting to a higher energy. The background current is measured before and after eachreading.This method is a combination of a collector-based and sample-biasing method; the transmissioncurrent is measured directly in the collector, while the reflection current is determined indirectly bydeducting it from the sample current. The method distinguishes fast electrons ( E se >
50 eV ) fromtrue secondary electrons ( E se <
50 eV ) by biasing the electrodes in the measurement setup. Thisrequires two separate measurements where the sample is first negatively biased and then positivelybiased.For a negative bias, the sample holder, retarding grid and collector are biased to −
50 V, 0 Vand 0 V, respectively. The negative bias repels fast and slow electrons from the sample on thereflection and transmission side. The transmission coefficient σ T ( E ) is determined by measuringthe transmission current with the retarding grid and collector and is given by: σ T ( E ) = I RG − + I C − I (2.1)where E is the electron energy of the primary electron, I is the primary beam current, I RG − isthe retarding grid current and I C − collector current. The minus-subscript indicates that the currentis measured under a negative sample bias. The total emission σ ( E ) , which is the sum of thereflection coefficient σ R ( E ) and transmission coefficient σ T ( E ) , is determined by measuring thesample current and is given by: σ ( E ) = I − I S − I (2.2)where I S − is the sample current. The reflection coefficient is then given by: σ R ( E ) = σ ( E ) − σ T ( E ) = I − I S − − I RG − − I C − I (2.3)An additional measurement with a positive biased sample can be performed to separate the fastelectrons from the slow ones. The sample holder, retarding grid and collector are biased to 50 V, 0 Vand 0 V, respectively. The positive voltage retracts the slow electrons to the sample, while allowingfast electrons ( E SE >
50 eV ) to escape. The retarding grid prevents tertiary electrons from thecollector wall (i.e. unwanted SEs induced within the setup) to flow back towards the sample. Theforward scattered electron coefficient η T ( E ) is determined by measuring the transmission currentwith the retarding grid and collector and is given by: η T ( E ) = I RG + + I C + I (2.4)where E is the electron energy of the primary electron, I is the primary beam current, I RG + isthe retarding grid current and I C + collector current, where the plus-subscript indicates a positivelybiased sample. Since σ T ( E ) = η T ( E ) + δ T ( E ) , the transmission secondary electron coefficient δ T ( E ) is given by δ T ( E ) = I RG − − I C − I − I RG + + I C + I (2.5)– 9 –he sample current I S + is again the sum of the reflection and transmission current. In this case η T ( E ) + η R ( E ) = I − I S + I (2.6)After substituting η T ( E ) , the backscattered electron coefficient η R ( E ) is given by: η R ( E ) = I − I S + − I RG + − I C + I (2.7)The reflection secondary electron coefficient δ R ( E ) can be determined by using the definition ofthe total emission coefficient: σ ( E ) = η R ( E ) + δ R ( E ) + η T ( E ) + δ T ( E ) , from which it followsthat δ R ( E ) = σ ( E ) − η T ( E ) − δ T ( E ) − η R ( E ) (2.8a) δ R ( E ) = I − I S − − I RG − − I C − I − I − I S + − I RG + − I C + I (2.8b)With eq. (2.5), (2.6), (2.7) and (2.8b), all relevant yields can be calculated from the measuredcurrents. In figure 4, a schematic drawing of a bi-layer membrane is given with all the currents that flowto and from the irradiated region. The flat side of the sample with the ALD alumina layer isfacing downwards in the transmission direction, while the window in the silicon substrate is facingupwards. The conductive TiN layer is deposited inside the window opening.The (secondary) electron yield curves as a function of the primary electron energy E are givenin figure 5 for bi-layer TiN/Al O membranes with thicknesses of 2.5/10, 5.7/25 and 5.7/50 nm,respectively. The total effective film thickness d is calculated with formula (1.6) and listed in table1 along with the transmission yield curve characteristics; the critical energy E c , max transmissionyield σ max T ( E max T ) and max transmission secondary electron yield δ max T ( E maxTSE ) .The reflection SEE coefficients are represented by the red curves in figure 5. For a bi-layermembrane, the contribution to the reflection yields is from the TiN layer. The back-scatteredelectron (BSE) coefficient η R ( E ) is close to zero for all three thicknesses. There are two factorsthat contribute to this low value. First, the BSE yield of membranes and foils are expected to belower in comparison with their bulk counterpart [34, 35]. Second, the window reduces the field ofview for BSEs and some will be recaptured.The reflection secondary electron (RSE) coefficent δ R ( E ) is below 1 and is lower than expected.The reduction in yield can again be attributed to recapture. The maximum reflection yield on abulk TiN sample can range from 1.4 to 2.8 for primary electron energy of 300 eV depending on thedeposition technique and conditions [36, 37].The transmission SEE coefficients are represented by the black curves in figure 5. Thetransmission side consists of Al O . The forward-scattered electron (FSE) coefficient η T ( E ) isthe fraction of the primary electron beam that penetrates through the membrane and retain energy– 10 – able 1 : Summary of important electron emission values of all composite membranes. The totaleffective thickness d is calculated with eq. (1.6).Type d Al O d TiN d Al O d σ max T E max T δ max T E maxTSE E c (nm) (nm) (nm) (nm) (keV) (keV) (keV)Bi-layer - 2.5 10 13.8 2.6 1.45 2.2 1.35 0.5Bi-layer - 5.7 25 33.6 2.1 2.55 1.7 2.15 1.0Bi-layer - 5.7 50 58.6 1.9 3.55 1.5 3.15 1.4Tri-layer 5 2.5 5 13.8 3.1 1.55 - - -Tri-layer 12.5 5.7 12.5 33.6 2.7 2.75 - - - E >
50 eV. Thin films become transparent for high-energetic electrons. As such, all primaryelectrons should end up in the collector, i.e. the FSE curve should approach 1 for high primaryelectron energy. However, the curves converge to 0.8 instead. The discrepancy can be attributed to(back)scattering of transmitted PEs on the retarding grid and the collector wall, which will inducetertiary currents that can lower the net transmission current. The effect of tertiary currents on thetransmitted fraction will be discussed in 4.2. In appendix B, a correction factor is estimated bytaking scattering events in the collector into account.The ’true’ transmission secondary electron (TSE) coefficient δ T ( E ) represents electrons with E se <
50 eV, which originates from the Al O layer within the escape depth. The initial rise ofthe TSE yield curve starts at the threshold energy E th . At this energy, the first (slow) secondaryelectrons emerges from the membrane in transmission. It is correlated to the critical energy E c forwhich 1% of the PEs are transmitted. These are defining yield curve characteristics that dependson the total effective thickness d of the membrane. Another characteristic is the maximum TSEyield δ max T obtained with PEs with electron energy E maxTSE . The thinnest membrane with a totaleffective thickness of 13 . .
35 keV). The maximum TSE
Figure 4 : The currents to and from a bi-layer membrane irradiated by an electron beam. TiNis sputtered into the window on the reflection side to provide a conductive path form the sampleholder to the irradiated region. – 11 – a) 2.5 / 10 nm (b) 5.7 / 25 nm(c) 5.7 / 50 nm
Figure 5 : Electron emission coefficients of a bi-layer membrane TiN/Al O with thicknesses d TiN / d Al O yield δ max T ( E maxTSE ) of the other membranes are listed in table 1.The total transmission coefficient σ T ( E ) is the sum of δ T ( E ) and η T ( E ) . In literature, thedistinction between δ T ( E ) and η T ( E ) is often not made. Unless specified specifically, usuallythe total transmission yield σ T ( E ) is given. The performance of a tynode can be expressed bythe maximum transmission yield σ max T ( E max T ) . The highest maximum transmission yield of 2.55(1 .
45 keV) was measured on a membrane with d = . In figure 6, a schematic drawing of a tri-layer membrane is shown. The titanium nitride layer issandwiched between two layers of alumina. The currents flowing to and from the irradiated regionsare indicated by the arrows.In figure 7a, the reflection σ R ( E ) and transmission coefficients σ T ( E ) of a bi-layer membraneTiN/Al O are compared to a tri-layer membrane Al O /TiN/Al O . The thicknesses of the layers– 12 – igure 6 : The currents to and from a tynode with a sandwiched TiN layer irradiated by an electronbeam. The reflection side is also covered with Al O , which protects the conductive layer duringthe fabrication process.for the two membranes are 2.5/10 nm and 5/2.5/5 nm, respectively, with a total effective thicknessof 13 . σ R ( E ) is significantly smaller for the bi-layer compared to the tri-layer, since the material of the emission surfaces are different. The reflection yield of TiN is lowerthan that for Al O . Therefore, a direct comparison is not useful.The transmission coefficient σ T ( E ) for both type of membrane has the same threshold energy E th . This shows that both membranes have a similar thickness and stopping power. However, themaximum transmission coefficient σ maxT of 3 . .
55 keV) is higher for the tri-layer membranecompared to the bi-layer yield of 2.6 (1 .
45 keV). The better performance is also observed for themembrane with d = . σ maxT of2.7 (2 .
75 keV) for the tri-layer compared to 2.1 (2 .
55 keV) for the bi-layer. Hence, encapsulatingthe conductive layer of TiN between two layers of Al O improves the transmission (secondary)electron yield in comparison with the bi-layer membrane. Before a more meaningful comparison can be made, additional data and corrections need to bemade. First, the reduction in yield due to reabsorption of SEs by the walls of the window openingneeds to be addressed. The aspect ratio of the window and the wall is ∼ a) (b) Figure 7 : Electron emission yield curves of a bi-layer membrane compared to a tri-layer membraneAl O /TiN/Al O with the same total effective thickness: (a) d = . d = . O emission surfaces provides more insight.Accordingly, both demands can be fulfilled by performing an additional measurement on thebi-layer membranes and combining the results of the two separate measurements. In the first,Al O /TiN/Al O -layer is facing downwards, so emission in transmission is unobstructed. In thesecond, the sample is turned over so that Al O /TiN/Al O )-layer is facing upwards. In this case,reflection SEE will be from Al O and correction is not needed. The transmission yield from thefirst measurement is then combined with the reflection yield of the second. The combined yieldcurves are shown in figure 8 for the bi-layer membranes with d = .
68 nm and d = . O membrane.The max reflection yield σ max R ( E max R ) is 3.3 (0 .
30 keV) and 3.7 (0 .
35 keV), respectively. This resultis close to the maximum reflection yield of an ALD Al O -film (12 . σ max R of 3.6 (0 . σ max R for the tri-layer membrane was outside themeasurement range. Also, a correction needs to be made to the reflection yield, since the windowopening is present on that side. If a correction factor of ∼
40 %, then the highest reflection yield is4.1 (0 .
35 keV) and 3.5 (0 .
75 keV), for d . d = . d = . d = . δ p and back-scattered (primary) electrons δ b : δ R = δ p + δ b [7]. In bulk samples (and thick films), a large contribution to RSE generationcomes from back-scattered electrons that dissipate energy when they leave from the interior. Inan experiment, where an aluminum target was irradiated with keV-electrons, the back-scatteredelectrons contributed close to 40% of the generated RSEs [34]. Also, backscattered electrons werefound to be 4.9 times as effective in generating SEs compared to incoming PEs. In thin films, the– 14 – igure 8 : The combined results obtained from two separate measurement on the same sample,where the flat surface is first facing downwards to obtain σ T , ↓ and then upwards to obtain σ R , ↑ .The combined results portray the electron emission characteristic of a flat Al O membrane with athickness of 13.8 nm and one with 33.6 nm.backscatter contribution δ b is negligible when R ( E ) (cid:29) d , since most will be transmitted throughthe film. The reflection yield curve or the bi-layer film with d = . δ p .A similar graph was found for thin Al films and Al bulk material by Kanter [34].As such, the thickness d = . O films. Thetransmission yield of the tri-layer is almost as large as the reflection yield of the film-on-bulksample, with σ max T = . σ max R = . δ max R of a thin film is alwayslarger than its TSE yield δ max T as shown by Ono et al using empirical formulas [38]. Thus, thetransmission yield of the tri-layer is already near its optimum. Reducing the thickness furtherwould decrease the interaction volume of the PEs with the material. Both the reflection and thetransmission yield would decrease below the optimum thickness. A further reduction of the filmthickness is warranted when the focus is on lowering the electron energy E max T . The transmitted fraction is a different name for the FSE coefficient η T . In early experiments, thetransmitted fraction was measured to determine electron-range relations, such as formula (1.5) [8].In figure 10a, the transmitted fraction of the bi-layer films are given. A correction factor of 0.2has been applied to η T ( E PE ) to account for tertiary currents in the semi-spherical collector (seeappendix B). In figure 10b, the energy is normalized by using the reduced initial energy E / E c .The critical energy E c is the energy for which 1% of the PEs are transmitted, which is relatedto the film thickness. This normalization was proposed by Kanter to determine the transmissioncurve of foils with different materials and thicknesses [39]. The transmission curve approachesa universal transmission curve for large film thickness ( d c ≤
20 µg / cm ) and is unique for each– 15 –aterial. Thinner films deviated from this universal curve as was shown for carbon foils. It issimilar to our result in figure 10b in which the curve for d = . (a) (b)(c) Figure 9 : Transmitted fraction as a function of (a) the primary electron energy, (b) the reducedinital energy E / E c and (c) the reduced thickness d / R of the membranes. The p-value for the threefilm thicknesses are 1.25, 1.55 and 1.88, respectively.The transmitted fraction can also be normalized by using the reduced film thickness d / R asshown in figure 9c. The transmitted fraction is given by: η T ( E ) = exp (cid:34) − . (cid:18) dR ( E , Z ) (cid:19) p ( E , Z ) (cid:35) (4.1)where E is the PE energy, d is the film thickness and R the range and p the transmission parameterrespectively [8]. Transmission curves normalized this way can be characterized by the transmissionparameter p . In 9c, the transmission characteristics of different p -values are plotted as well. Lighterelements have a transmission characteristic similar to the curve with p ≈
2, while heavier elementshave a curve similar to p ≈ .
5. The p -value is constant for thick films and is 1.9 for Al O . In– 16 –gure 9c, the reduced film thickness d / R is determined for each film thickness by using eq. (1.5)to calculate the range for each E . The transmission parameter is determined by superimposingcurves calculated with (B.1) onto the normalized measurement data. Again, the thicker film with d = . p ≈ .
8. However, thep-value and characteristics deviates for thinner films with d < d c .Figure 9b and 9c shows the same effect: Thin films with d < d c have different transmissioncharacteristic. This is relevant for tynodes, since they are operated with sub-2-keV electrons. Thetransmission characteristic of the film with d = . The semi-empirical formula (1.3) is used to predict universal reflection secondary electron yieldcurves by normalizing the yield and energy [5, 30]. Different models have been opted that havevarying degrees of success in predicting the universal RSE yield curve for different materials.Though, it is beyond the scope of this paper. A recent review paper discuss the models moreextensively [31].More relevant for this paper is the question whether a universal transmission secondary electronyield can be found. The TSE yield curves are shown in figure 10a. The normalization is done forboth the TSE yield δ T / δ max T and the energy E / E max0 and is shown in figure 10b. The normalizedcurves have the same resemblance, but differs slightly. A similar result was found for carbon foilsby HÃűlzl et al [35]. They performed a similar experiment in which they treated η R , T ( E ) and δ R , T ( E ) separately. The results hints towards the existence of a universal TSE curve.If we look at the semi-empirical formula (1.4), the escape probability of SE at depth x isgiven by formula (1.2). The escape probability f ( x ) is the same regardless of the film thickness,since the material and the surface conditions are the same. The difference in δ T ( E ) is solelydue to the energy loss function dE / dx . There are numerous theoretical models for the energytransfer profiles and it is beyond the scope of this paper to discuss them. However, the shape of theenergy transfer profile is the same when they are normalized by using the reduced thickness d / R [40]. Therefore, since a universal energy transfer profile exist, a universal transmission secondaryelectron yield curve should exist as well. This is true for films with d > d c , as we have seen thatthe transmission characteristics of PEs converge to the same curve for thick films. However, thetransmission characteristics deviates for thinner films with d < d c and hence the energy transferprofile as well.Since η T ( E ) and δ T ( E ) are correlated, a logical representation of the data would be plottingthe reduced yield δ / δ max T versus the transmitted fraction η T as shown in figure 11. The maximumTSEY coincides with a transmitted fraction of approximately 0.4 to 0.5. Transmitted electrons stillpossess a considerable amount of energy: E x ( x = R ) ≈ (0.3 to 0.4) E [40]. Therefore, they shouldnot be neglected when designing a detector, because they can induce tertiary currents and feedbacksignals. – 17 – a) (b) Figure 10 : (a) TSEY (b) Normalized TSE yield curve
Figure 11 : Normalized yield vs. transmitted fraction
We have successfully constructed multilayered Al O /TiN freestanding films that can be usedas tynodes in TiPC. Two types of films have been made, a bilayer TiN/Al O and a tri-layerAl O /TiN/Al O . The Tri-layer film has the conductive TiN layer encapsulated in order to protectit during manufacturing and to improve the reliability of the manufacturing process. The highesttransmission yield was attained for the thinnest tri-layer membrane which have a yield of 3.1(1.55 keV). The (total effective) thickness of 13 . d > d c , a universal transmission yield curve seems to exist. However, for thinner films with d < d c the energy transfer profile differs and universality is not to be expected.Although the transmission yield can be further improved, the tri-layer membrane can be usedto build a rudimentary/prototype TiPC. Improvement to the transmission yield can be achievedby surface termination, such as caesiation or hydrogen-termination. Also, other materials can beconsidered as emission material, such as MgO. Magnesia has a higher reflection yield comparedto alumina and can be deposited with ALD as well. The same fabrication process can be used aspresented in this paper, but replacing the ALD Al O with ALD MgO instead. Lastly, the activesurface area of tynodes can be increased by forming meta-materials with ALD Al O /TiN/Al O .This can improve the collection efficiency of the Timed Photon Counter. Acknowledgments
We would like to thank H. Akthar, C. Hansson, S. Tao and J. Smedley for their contribution to theMEMBrane project. We are also thankful for the technical support we received from O. v. Petten(Nikhef), J. v.d. Cingel, H. v.d. Linden and C.T.H. Heerkens on the experimental setup. We aregrateful to the Else Kooi Lab that provided the training and facilities to manufacture the films. Thiswork is supported by the ERC-Advanced Grant 2012 MEMBrane 320764.
References [1] Hamamatsu Photonics K.K.,
Photomultiplier tubes
IEEE Journal on Selected Topics in Quantum Electronics , vol. 20, no. 6,2014.[3] H. van der Graaf, H. Akhtar, N. Budko, H. W. Chan, C. W. Hagen, C. C. Hansson, G. Nützel, S. D.Pinto, V. Prodanović, B. Raftari, P. M. Sarro, J. Sinsheimer, J. Smedley, S. Tao, A. M. Theulings, andK. Vuik, “The Tynode: A new vacuum electron multiplier,”
Nuclear Instruments and Methods inPhysics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment , vol.847, no. June 2016, pp. 148–161, 2017.[4] H. Bruining,
Physics and Applications of Secondary Electron Emission , 1st ed. McGraw-Hill BookCompany, 1954. [Online]. Available: https://books.google.nl/books?id=YrgEAAAAMAAJ[5] A. Dekker, “Secondary Electron Emission,”
Solid State Physics , vol. Volume 6, pp. 251–311, 1958.[Online]. Available: https://doi.org/10.1016/S0081-1947(08)60728-6[6] J. E. Yater and a. Shih, “Secondary electron emission characteristics of single-crystal andpolycrystalline diamond,”
Journal of Applied Physics , vol. 87, no. 11, p. 8103, 2000. [Online].Available: http://link.aip.org/link/JAPIAU/v87/i11/p8103/s1{&}Agg=doi[7] K. Kanaya and H. Kawakatsu, “Secondary electron emission due to primary and backscattered,”
Journal of Applied Physics , vol. 5, no. 1957, 1972. – 19 –
8] H. J. Fitting, “Transmission, energy distribution, and SE excitation of fast electrons in thin solidfilms,”
Physica Status Solidi (a) , vol. 26, no. 2, pp. 525–535, 1974.[9] E. J. Sternglass, “High-Speed Electron Multiplication by Transmission Secondary ElectronEmission,”
Review of Scientific Instruments , vol. 26, no. 12, p. 1202, 1955. [Online]. Available:http://scitation.aip.org/content/aip/journal/rsi/26/12/10.1063/1.1715223[10] E. J. Sternglass and M. M. Wachtel, “Transmission Secondary Electron Multiplication forHigh-Speed Pulse Counting,”
IRE Transactions on Nuclear Science , vol. 3, no. 4, pp. 29–32, nov1956. [Online]. Available: http://ieeexplore.ieee.org/document/4315542/[11] W. W. Sapp and E. J. Sternglass, “Planar Dynode Multipliers for High-Speed Counting,”
IEEETransactions on Nuclear Science , vol. 11, no. 3, pp. 108–113, 1964. [Online]. Available:http://ieeexplore.ieee.org/document/4323412/[12] J. Llacer and E. L. Garwin, “ElectronâĂŘPhonon Interaction in Alkali Halides. II. TransmissionSecondary Emission from Alkali Halides,”
Journal of Applied Physics , vol. 40, no. 7, pp. 2776–2792,jun 1969. [Online]. Available: http://aip.scitation.org/doi/10.1063/1.1658076[13] M. Hagino, S. Yoshizaki, M. Kinoshita, and R. Nishida, “Caesium Activated CsI Transmission-typeSecondary Emission Dynode,”
Advances in Electronics and Electron Physics , 1972.[14] R. U. Martinelli, “Reflection and Transmission Secondary Emission From Silicon,”
Applied PhysicsLetters , vol. 17, no. 8, p. 313, 1970. [Online]. Available:http://scitation.aip.org/content/aip/journal/apl/17/8/10.1063/1.1653414[15] J. Yater, A. Shih, J. Butler, and P. Pehrsson, “Electron transmission studies of diamond films,”
AppliedSurface Science , vol. 191, no. 1-4, pp. 52–60, may 2002. [Online]. Available:http://linkinghub.elsevier.com/retrieve/pii/S0169433202001496[16] J. E. Yater, A. Shih, J. E. Butler, and P. E. Pehrsson, “Transmission of low-energy electrons inboron-doped nanocrystalline diamond films,”
Journal of Applied Physics , vol. 93, no. 5, pp.3082–3089, 2003.[17] ——, “Electron transport mechanisms in thin boron-doped diamond films,”
Journal of AppliedPhysics , vol. 96, no. 1, pp. 446–453, 2004.[18] J. E. Yater, J. L. Shaw, K. L. Jensen, T. Feygelson, R. E. Myers, B. B. Pate, and J. E. Butler,“Secondary electron amplification using single-crystal CVD diamond film,”
Diamond and RelatedMaterials , vol. 20, no. 5-6, pp. 798–802, 2011. [Online]. Available:http://dx.doi.org/10.1016/j.diamond.2011.03.040[19] S. X. Tao, H. W. Chan, and H. Van Der Graaf, “Secondary electron emission materials fortransmission dynodes in novel photomultipliers: A review,”
Materials , vol. 9, no. 12, 2016.[20] V. Prodanovic, H. W. Chan, J. Smedley, A. Theulings, and S. Tao, “Optimization of silicon-richsilicon nitride films for electron multiplication in timed photon counters,”
Procedia Engineering , vol.120, pp. 1111–1114, 2015. [Online]. Available: http://dx.doi.org/10.1016/j.proeng.2015.08.804[21] V. Prodanović, H. W. Chan, H. V. Graaf, and P. M. Sarro, “Ultra-thin alumina and silicon nitrideMEMS fabricated membranes for the electron multiplication,”
Nanotechnology , vol. 29, no. 15, 2018.[22] A. Chvyreva and A. J. M. Pemen, “Experimental investigation of electron emission from dielectricsurfaces due to primary electron beam: a review,”
IEEE Transactions on Dielectrics and ElectricalInsulation , vol. 21, no. 5, pp. 2274–2282, oct 2014. [Online]. Available:http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6927357 – 20 –
23] N. R. Whetten, “Cleavage in high vacuums of alkali halide single crystals - Secondary electronemission,”
Journal of Applied Physics , vol. 35, no. 11, pp. 3279–3282, 1964.[24] K. Handel, A. Jensen, and M. Siedband, “A two electron gun technique for the measurement ofsecondary emission characteristics of a variety of materials,”
IEEE Transactions on Electron Devices ,vol. ED-13, no. 6, pp. 525–528, jun 1966. [Online]. Available:http://ieeexplore.ieee.org/document/1474315/[25] N. Balcon, D. Payan, M. Belhaj, T. Tondu, and V. Inguimbert, “Secondary electron emission on spacematerials: Evaluation of the total secondary electron yield from surface potential measurements,”
IEEE Transactions on Plasma Science , vol. 40, no. 2 PART 1, pp. 282–290, 2012.[26] S. Logothetidis, E. I. Meletis, G. Stergioudis, and A. A. Adjaottor, “Room temperature oxidationbehavior of TiN thin films,”
Thin Solid Films , vol. 338, no. 1-2, pp. 304–313, 1999.[27] D. M. Taylor and Q. H. Mehdi, “Electron-beam-induced conduction in SiO2 thin films,”
Journal ofPhysics D: Applied Physics , vol. 12, no. 12, pp. 2253–2267, dec 1979. [Online]. Available:https://iopscience.iop.org/article/10.1088/0022-3727/12/12/028[28] J. Lee, “Thickness effect on secondary electron emission of MgO layers,”
Applied Surface Science ,vol. 174, no. 1, pp. 62–69, 2001. [Online]. Available:http://linkinghub.elsevier.com/retrieve/pii/S0169433201000150[29] A. E. Vladar, M. T. Postek, and R. Vane, “Proceedings of Active monitoring and control of electronbeam induced contamination,”
Proceedings of SPIE - The International Society for OpticalEngineering
Journal of AppliedPhysics , vol. 46, no. 8, pp. 3347–3351, 1975.[31] P. Tolias, “On secondary electron emission and its semi-empirical description,”
Plasma Physics andControlled Fusion , vol. 56, no. 12, p. 123002, 2014. [Online]. Available: http://stacks.iop.org/0741-3335/56/i=12/a=123002?key=crossref.8781cd8263ae6d451343cd8e5c4eb570[32] F. A. Lukiyanov, E. I. Rau, and R. A. Sennov, “Depth range of primary electrons, electron beambroadening, and spatial resolution in electron-beam studies,”
Bulletin of the Russian Academy ofSciences: Physics
Journal of Electron Spectroscopy and Related Phenomena ,vol. 159, no. 1-3, pp. 46–52, 2007.[34] H. Kanter, “Contribution of Backscattered Electrons to Secondary Electron Formation,”
PhysicalReview , vol. 121, no. 3, pp. 681–684, feb 1961. [Online]. Available:https://link.aps.org/doi/10.1103/PhysRev.121.681[35] J. Hölzl and K. Jacobi, “Sekundärelektronen-ausbeute an dünnen kohlenstoffolien in transmission undreflexion,”
Surface Science , vol. 14, no. 2, pp. 351–360, 1969.[36] P. He, H. C. Hseuh, R. Todd, B. Henrist, N. Hilleret, F. L. Pimpec, R. E. Kirby, M. Pivi, S. Kato, andM. Nishiwaki, “Secondary Electron Emission Measurements for Tin Coating on the Stainless Steel ofSns Accumulator Ring Vacuum Chamber,”
Epac 2004 , no. July, pp. 5–8, 2004. – 21 –
37] D. Wang, Y. He, and W. Cui, “Secondary electron emission characteristics of TiN coatings producedby RF magnetron sputtering,”
Journal of Applied Physics , vol. 124, no. 5, 2018.[38] S. Ono and K. Kanaya, “The energy dependence of secondary emission based on the range-energyretardation power formula,”
Journal of Physics D: Applied Physics , vol. 12, no. 4, pp. 619–632, apr1979. [Online]. Available: https://iopscience.iop.org/article/10.1088/0022-3727/12/4/019[39] H. Kanter, “Electron Scattering by Thin Foils for Energies Below 10 kev,”
Physical Review , vol. 121,no. 2, pp. 461–471, jan 1961. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRev.121.461[40] H. J. Fitting, “Six laws of low-energy electron scattering in solids,”
Journal of Electron Spectroscopyand Related Phenomena , vol. 136, no. 3, pp. 265–272, 2004.
Appendices
A Sample geometry correction
Reabsorption by the window opening is estimated by comparing two measurements on a p-typesilicon membrane with a thickness of 39 nm. The membrane is symmetric in this case except for thewindow. The width, height and aspect ratio are the same as the other membranes presented in thispaper. In figure 12a, the reflection and transmission curves of two separate measurements are given.The first measurement, with the window facing upwards, shows a reduction in reflection yield, whilethe transmission yield is unaffected. In the second measurement, the window is facing downwardsand the result shows a reduction in the transmission yield, while the reflection yield is unaffected.The ratio between the reduced and unaffected yields is given in figure 12b. Reabsorption decreasesthe reflection yield by 35 to 45% and the transmission yield by 15 to 30%. A correction can beapplied to the obtained results, but the correction factor depends on the beam energy and only holdsfor these specific samples.
B Measurement setup correction
A transmitted PE can cause tertiary currents on the retarding grid and collector, when the setup isoperated with a positive sample bias. The sample holder and collector are positively biased withrespect to the retarding grid. First, a transmitted PE that lands on the grid can generate tertiaryelectrons that will either flow to the collector or the sample holder due to the bias. This induces atertiary current I tertiary from the grid. The tertiary current from the grid to the collector will have azero net effect on the transmitted fraction, since η T ( E ) = ( I RG + − I tertiary ) + ( I C + + I tertiary ).Second, a transmitted PE has a chance to backscatter on the wall of the collector. Tertiaryelectrons will be generated as well, but they will be recaptured by the collector due to the bias.The net current from the collector to the sample holder will only consist of the backscatter current I backscatter . – 22 – a)(b) Figure 12 : (a) The influence of reabsorption by the window opening on secondary electron emissionof a silicon membrane on a SOI substrate. (b) The ratio between the obstructed and unobstructedyield for reflection and transmission yield.Overall, the measured transmitted fraction is given by: η T , measured ( E ) = I T − I backscatter − I tertiary I (B.1a) = I T I − I backscatter + I tertiary I (B.1b) = η T , true ( E )) − I backscatter + I tertiary I (B.1c) η T , measured ( E ) = η T , true ( E ) + α ( E ) (B.1d)where I T is the (true) transmission current in nA, η T , true ( E ) is the true transmitted fraction, I backscatter is the backscatter current from the collector to the sample holder in nA, I tertiary the tertiary current– 23 –rom the grid to the sample holder in nA and α ( E ) a correction factor. The tertiary current is givenby: I backscatter = I γδ R , grid (B.2)where I is the primary current, γ the opacity of the retarding grid mesh and δ R , grid the reflectionyield of the grid mesh material. The backscatter current is given by: I tertiary = I ( − γ ) ε θ η R , col (B.3)where I is the primary current, γ the opacity of the retarding grid mesh, ε θ the backscatter angleefficiency and η R , col the backscatter yield of the collector material. The correction factor α can beestimated by: α = I backscatter + I tertiary I = I γδ R , grid + I ( γ ) ε θ η R col I (B.4a) = γδ R , grid + ( − γ ) ε θ η R , col (B.4b) = . × . + ( − . ) × . × . ≈ .
22 (B.4c)with the assumption that δ R , grid = . η R , col = .
22 for copper and a mesh opacity γ = .
1. This is a rough estimate, since the exact scattering angles and multiple scattering eventsare not taken into consideration. Also, the correction factor depends on the primary electron energy E0