Mechanical Meta-material of Multi-layered Al 2 O 3 /TiN/Al 2 O 3 film as Large-surface Transmission Dynode
H.W. Chan, V. Prodanović, T. ten Bruggencate, C.W. Hagen, P.M. Sarro, H. van der Graaf
PPrepared for submission to JINST
Mechanical Meta-material of Multi-layered Al O /TiN/Al O film as Large-surface Transmission Dynode H.W. Chan, a , b , V. Prodanović, a , b T. ten Bruggencate 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: We present a novel manner in which to fabricate tri-layer transmission dynodes (tynodes).These tynodes are self-supporting and reinforced by silicon-oxide. This allows us to span a largerarea, which is necessary for the development of the Timed Photon Counter. The performance ofthese tynodes is measured using a scanning electron microscope setup.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 The Timed Photon Counter (TiPC) is a novel vacuum electron multiplier that has ultra-thin trans-mission dynodes (tynodes) as multiplication stages [1]. The detection principle is similar to aphotomultiplier tube (PMT), namely photoemission from a photocathode and subsequent (sec-ondary) electron multiplication in vacuum. In a PMT, a photoelectron is multiplied within a seriesof (reflective) dynodes [2]. The dynodes are carefully aligned in a sequence, where the electricpotential is step-wisely increased between each dynode. Under influence of the electric field, theSEs are accelerated and directed from dynode to dynode. Their number increases until they reachthe anode, which will provide an electrical signal as read-out. PMTs belong to one of the mostsensitive photon detectors due to its high gain and low noise. However, there are some disadvantagesdue to the complex arrangement of the (reflective) dynodes. First, the non-uniform electron pathsbetween each dynode limits the timing resolution to nanoseconds. Second, the electron paths areeasily perturbed by external magnetic fields. And lastly, the device is large, fragile and expensiveto manufacture.TiPC will improve upon PMTs in terms of timing and spatial resolution by using tynodes. Atynode is an ultra-thin membrane that allows for transmission secondary electron emission (TSEE):an incoming primary electron (PE) will generate multiple transmission secondary electrons (TSE)within a membrane as the PE transmits through it. This allows tynodes to be closely stacked on topof each other, which benefits the timing resolution. The more homogenous and stronger electric– 1 –eld between the tynodes will reduce the spread in arrival time of TSEs. Also, the TSEs are lesssusceptible to external magnetic field due to the increased electric field strength. Moreover, 2Dspatial information can be acquired with the planar tynodes by using a CMOS pixelchip as read-out.The performance of a tynode can be characterized by the maximum transmission yield σ max T obtained with primary electron energy E max T : σ max T ( E max T ) . In a detector with multiple tynodes, theformer will determine the overall gain, while the latter the required operating voltage. The requiredPE energy E max T is correlated to the film thickness and the range of the electron in the film. Afterall, only SEs generated near the exit surface of the membrane has a chance to escape. As such,the thickness of the film is a parameter that needs to be optimized. In the past, several types oftransmission dynodes with high transmission yields (TEY) have been reported [3]. Unfortunately,they were impractical due to the required PE energy to achieve it. For instance, a TEY of 27 (9 keV)was reported for a caesium-activated CsI film deposited on Al/Al O film by Hagino et al. [4]. Adevice with several of these tynodes will require a very large operating voltage.Our group used Micro-Electro-Mechanical System (MEMS) technology to fabricate ultra-thinmembranes, which were suspended within a supporting mesh with small circular windows with adiameter of 30 µm [5]. The windows are arranged in a 64x64 array with a pitch of 55 µm to matchthe spacing of the pixel pads on a TimePix chip. An aligned tynode stack will form a channel inwhich electron multiplication will take place and be read out by individual pixels as shown in figure1 . Two types of material was used to form the ultra-thin membrane: Low-pressure-chemical-vapor-deposition (LPCVD) silicon nitride and atomic-layer-deposition (ALD) aluminum oxide.The former has a transmission yield of 1.6 (2.8 keV) for a membrane with a thickness d =
40 nm,while the latter has a transmission yield of 2.6 (1.45 keV) for a membrane with d =
10 nm. Onboth tynode, a titanium nitride (TiN) layer is deposited on one side to provide in-plane conductivity.In a different process, the TiN layer was encapsulated within two layers of Al O to improve the Figure 1 : Schematic drawing of TiPC withtynodes comprised of a small window array
Figure 2 : Schematic drawing of TiPC withmetamaterial tynodes– 2 –eliability of the fabrication process [6]. This tri-layer film had a transmission yield of 3.1 (1.55keV).The supporting mesh is a necessity in the design of the tynode array due to the fragility ofthe ultra-thin films, but does not contribute to electron multiplication. Photoelectrons that areemitted from the photocathode above the mesh will not be detected. The collection efficiencyof TiPC is therefore proportional to the active surface of the tynode, which can be estimated bythe ratio between the surface of the small windows and the supporting mesh. For an array ofwindows with a diameter of 30 µm and a displacement of 55 µm, the active surface is only 23 . . O film on a 3D honeycomb mold, which is removed afterwards by plasma etching. Theimproved properties are ascribed to the extruded (semi-open hollow) ribs which provide additionalmechanical strength.The metamaterial can be functionalized as a tynode by adding a conductive layer, such astitanium nitride, to the ALD Al O film [5, 6]. A tynode formed as a metamaterial has the samethickness continuously. Incoming electrons can generate SEs on any part of the structure. Asa result, the effective area of these tynodes will be nearly 100% and the alignment precision inthe tynode stack will be less stringent. In addition, the honeycomb-shaped domes will provide afocusing effect. Also, there is no longer a risk of any charge-up effect, since the thick dielectricmesh is no longer present. And lastly, the fabrication process is less complex in comparison. Infigure 2, a schematic drawing of a TiPC detector with metamaterial tynodes have been drawn, whichcompared to figure 1 has a larger active surface for electron multiplication.In this paper, metamaterials with two different patterns are presented: one with a hexagon/honeycombpattern and one with an octagon pattern. The latter is designed to match the pitch of the pixel pads– 3 –f a TimePix chip. The secondary electron emission film is a tri-layer Al O /TiN/Al O film de-posited by atomic-layer-deposition (ALD). The electron yield is determined with a collector-basedmethod within a scanning electron microscope. Two periodic cellular patterns are considered in this paper: hexagonal/honeycomb and octagonal.The first is a hexagonal pattern similar to the one reported by Davami et al [8]. They have shown thatmetamaterials with this pattern have increased strength, flexural stiffness, rigidity and resistance todeformation. The free-standing plate material of alumina can have a length of up to 2 cm and areextremely light and resilient. Also, it has been demonstrated that it has shape-recovering properties.The bending stiffness is ascribed to the hexagon pattern in which any plane perpendicular to theplate is intersected by a vertical wall. In addition, any cracks have to propagate along the triangularlattice, which makes these materials more robust. They derived a design rule for new patterns: anyplane perpendicular to the plate needs be intersected by a vertical wall.
Figure 3 : COMSOL simulation of a TiPC detector with two tynode stages. For the simulation,two modules of COMSOL was used: AC/DC and Charged Particle Tracing (CPT). The first moduleis used to simulate the static electric field. ∆ U = igure 4 : The metamaterials are suspended within a silicon frame with a dimension of 20 mm by20 mm. For the hexagonal pattern, 16 windows with a surface of approximately 1 mm are openedin which the metamaterial film is suspended. For the octagonal pattern, the film is suspended in asingle window with a surface of approximately 12 mm .With this rule in mind, the second pattern is an octagonal pattern which is designed towork in conjunction with a TimePix chip. First, the square lattice of the pattern matches thepitch/displacement of the pixel pads on a TimePix chip. Second, the octagonal-shaped domes havea focusing effect on SEs, which guide the SEs from each tynode in a ’channel’ above the individualpixel pads. This effect is also needed to focus the SEs onto the collection surface of the pixels,which has a smaller dimension than the spacing of a pixel.The focusing effect of these octagonal-shaped cells have been simulated with COMSOL (figure3). A cell with a height of 5 µm is already sufficient to focus SEs, but the focusing point can betailored by varying the height of the unit cells. The simulation also shows that PEs and SEs thatenter the small square pattern in the corner of each octagonal are ’lost’, since they land next to thepixel pads. This needs to be taken into consideration when estimating the collection efficiency.The dimensions of the cells are chosen to match the pitch of the pixels and is shown in figure 4for both patterns. The rib height and width are 5 µm, which is relatively shallow. Incoming electronscan still reach the bottom of the rib and contribute to electron multiplication. Increasing the heightfurther, the chance for incoming electrons to reach the bottom of the trench will decrease, sinceonly electrons that enter perpendicularly will manage to do so. The unit diameter is 50 µm and hasa pitch of 55 µm (which is the same as the pitch between pixels). The windows are differently sizedand are designed to open arrays with 16x16 or 64x64 unit cells. An additional margin is added toaccount for the slope due to the etching process. The windows have a width of (0.88+0.74) mm and(3.52+0.74) mm with a surface area of 2.64 and 18.2 mm , respectively.The emission film is a tri-layer Al O composite with a thickness of 10/5/15 nm. This film isfunctionalized as a SEE membrane by encapsulating a TiN layer, which is needed to sustain electronemission for prolonged operation [6]. – 5 – a) 3D silicon mold (b) ALD of tri-layer film onto the silicon mold(c) Release of the metamaterial Figure 5 : Flow chart a 4-inch p-type (5-10 Ω cm) wafer with a thickness of 500 ±
15 µm is used as substrate. The wafersare cleaned with a standard cleaning procedure before a zeroth layer is added with markers forcontact alignment. The sequence of the cleaning procedure is as follows: a plasma oxygen etch, aHNO 100% bath for 10 min, a demineralized water rinse for 5 min, a HNO 65% bath at 110 ◦ C for10 min and another water rinse.The process can be divided in three parts. In the first part, a 3D mold is etched within thesilicon substrate (figure 5a). A 3-µm-thick photoresist (PR) layer is used as a masking layer fora deep-reactive-ion-etch (DRIE) in a Rapier Omega i2L DRIE etcher. The hexagon pattern istransferred to the PR and trenches with a depth of 5 µm are etched into the silicon substrate. Thewafers are then cleaned with oxygen plasma to remove residual polymers from the DRIE processfollowed by a standard cleaning procedure. Afterwards, the wafers are put in an oven at 1100 ◦ Cto wet thermally grow a silicon dioxide layer with a thickness of 500 nm. This layer will act as asacrificial and stopping layer.In the second part, the tri-layer film material is deposited conformally on the mold (figure5b). First, a layer of Al O is deposited by atomic-layer-deposition (ALD) in a thermal ALD ASMF-120 reactor at 300 ◦ C using trimethyl-aluminum (TMA) and water as a precursor and reactant,respectively. A strip of the newly deposited layer is removed from the edge of each die to createa contact point for the next layer to the silicon substrate. First, the Al O layer is plasma etchedin an Omega Trikon plasma etcher. Then, the silicon dioxide layer is removed with a plasma etch– 6 –n a Drytek plasma etcher. The wafers are cleaned afterwards using a new procedure omitting the’fuming’ HNO 65% bath. Next, a layer of ALD titanium nitride is deposited in an Ultratech Fiji G2using titanium chloride (TiCl4) as precursor and nitrogen plasma as reactant at 250 ◦ C. The thirdlayer of Al O is then deposited with the same recipe as the first layer in the same reactor.In the third part, the metamaterial will be released by opening windows in the silicon substrate(figure 5c). First, a plasma-enhanced chemical vapor deposition (PECVD) oxide layer with athickness of 1 µm is deposited on the front side of the wafer in a Novellus Concept One system.This oxide layer protects the tri-layer film from subsequent etching steps. The backside of the waferis stripped by using two plasma etches. The ALD TiN and ALD Al O layers are stripped in theOmega Trikon plasma etcher, while the thermal oxide is removed in the Drytek plasma etcher. Once,the backside is stripped and cleaned, a PECVD oxide layer with a thickness of 5 µm is depositedon the backside of the wafer in the Novellus Concept One system. This PECVD oxide layer is themasking layer for DRIE. The pattern of the window openings and scribe lines are transferred usingPR with a thickness of 3 . The transmission secondary electron yield is measured with a collector-based method within ascanning electron microscope (SEM) as shown in figure 6a. A more thorough description of themethod is given in ref [6]. First the electron beam current is measured within a small Faraday cupnext to the sample. The beam is then moved to the window in the sample in which the ultra-thin filmis suspended. During the measurement, the beam is scanned over the surface in a zigzag patternusing the image acquisition mode of the SEM. This mode lowers the electron dose per unit surface,which mitigates charge-up effects and/or the built-up of surface contamination. The sample holder,mesh grid and collector are connected to Keithley 2450 source meters via a feedthrough into thevacuum chamber. The transmission current is measured directly within the collector, while thereflection current is determined indirectly using the sample current. The sample is either negativelybiased at −
50 V or positively at +
50 V. The former setting is used to determine the total reflectionand transmission yield, while the latter is used to measure the back- and forward scattered electrons.– 7 – a) (b)
Figure 6 : Experimental setup. (a) schematic drawing of the collector. (b) The copper collectorwith one of the octagonal metamaterial film mounted in the sample holder.By subtracting the results, the ’true’ reflection and transmission secondary electron yields can befound. The PE energy range from 0.3 âĂŞ 10 keV with a beam current of 0.06 to 0.54 nA. Thecurrent depends on the PE energy and is measured for each energy. The following SEM settingsare used during the measurement: a dwell time of 1 µs, a magnification of 50X,a Half-full-window(HFW) of 2 .
56 mm and a resolution of 1024 x 884. Using these settings, the area of the irradiatedsurface can be estimated, which in this case is 5 .
66 mm . The obtained yields are averaged over thesurface and over time, i.e. the yield is calculated from multiple frames. A transmission secondary electron yield map, as a function of the coordinates of the surface ofthe membrane, can be obtained during a (single) SEM image acquisition. A yield map will showthe difference in electron emission across the surface. The method is described in appendix A andoperates on the same principles as SEM image construction. The SEM settings are chosen such thatthe Keithley 2450 sourcemeters are able to map the measured current (as of function of time) to thepixels in the SEM image.This method require a much slower scan speed compared to the methodpresented in section 4.1. Also, only one image frame is acquired. A dwell time of 1 ms is chosen,which is a compromise between speed and accuracy. A larger dwell time might cause charge-upeffects and/or surface contamination. The resolution of the image is 512 x 442 acquired with adwell-, line- and frame time of 1 ms, 560 ms and 4 . . .
29 nA, which have the highest transmissionyield for the membranes considered in this paper. The magnification is 500X or 1000X, whichshows the difference in yield across the 3D structure of the metamaterial more clearly.– 8 –
Results & discussion
SEM images of the front- and backside of a metamaterial film with a hexagonal pattern is shown(figure 7). The contrast in the image is due to the 3D structure of the film. In figure 7a, the ribsextrude into the plane and behaves as ’trenches’ from which it is more difficult for SEs to escape.Therefore they appear darker. In figure 7b, the backside of the film is shown. In this case the ribsprotrude out of plane and appear brighter. The 3D structure of the film is more apparent in figure8, which shows a curled up film showing both sides. In figure 8, SEM images of the metamaterialwith the octagonal pattern is shown. The black dots on the surface are residues from the fabricationprocess. On a different sample, a close-up of a broken film shows flakey residues on the film andthe ribs (figure 10). These are residues of the HF vapor etch. Also, in figure 10b, the etch lineof the DRIE process is indented in the ribs. The DRIE process is a cyclic process of etching andpassivation. For the samples in this paper, 8 cycles were used to etch 5 . The metamaterial films have been measured with the unit cells/domes facing up- and downwards.Upward facing will have a more uniform response, since the emission surface is flat. Downwardfacing will have the focusing effect. Depending on the application, one might be more preferredthan the other.The maximum transmission yield of the octagon pattern with a window size of 12 . is2 . ( .
15 keV ) . Placed upside down, 1 . ( .
15 keV ) . Critical energy of approximately 1 . is 1 . ( .
05 keV ) . Placed upside down1 . ( .
05 keV ) . The lower yield of the hexagon is due to the window size, which recapture someof the emitted electrons and lowers the yield. The octagon sample has an almost open field of view.The effect is also clearly shown in the reflection yield, when the window is facing upwards. Thereflection yield is much lower.Compared to the bi-layer films in [5], which had similar thickness: 2 . ( .
55 keV ) with a criticalenergy of 1 keV. Compared to the tri-layer films in [5]: 2 . ( .
75 keV ) The membranes presented here are slightly thicker in comparison with the flat films. Theslight shift in critical energy E c of 150 eV and also the max peak of 400 −
600 eV indicates it.– 9 – a) Front (b) Back
Figure 7 : SEM images of a metamaterial film with a hexagonal pattern. (a) The contrast betweenthe active area and the window frame is due to the transparency of the film for 5 keV electrons.(b) On the backside, the ribs protrude outwards and appear brighter since SEs are generated closertowards the SE detector of the SEM. On the edge, the ribs disappear into the silicon substrate.
Figure 8 : A SEM image of a broken film curled up after release. It clearly shows the ribs of thehoneycombs on the front- and backside. – 10 – a) Front (b) Back
Figure 9 : SEM images of a metamaterial film with an octagonal pattern. (a) (b)
Figure 10 : SEM images of a broken metamaterial film with an octagonal pattern. The etch linesof the DRIE process is visible on the ribs. (a)The difference in thickness can stem from either: (1) production parameters or (2) residues fromprocessing (black dots on the surface/ crystals on the surface). The transmission yield is comparableto the bi-layer, but lower than the tri-layer.
The active surface of a tynode can be determined by measuring transmission yield as function ofthe coordinates of a tynode. In 13, the response of an octagon tynode is shown. The yield response– 11 – igure 11 : Electron yield curves of a hexagonal metamaterial film with a thickness of 15 / /
10 nm
Figure 12 : Electron yield curves of an octagonal metamaterial film with a thickness of 15 / /
10 nmis higher than one on the entire surface, thus we can assume that the active area is 100%. Thevariation of the yield across the surface is due to the vertical walls, which has a higher transmissionyield response due to the entry angle into the film. It has been shown that more SEs are generatedfor PEs that enters an angle. When placed upside down, the situation is slightly different. There isno focusing effect (except within the ribs). The response will also be different and a more uniformemission is expected, since the emission surface is mostly flat. Within the ribs, the transmissionyield is lower due to recapture of SEs in the rib with a 1:1 height/width ratio.It is difficult to predict what the collection efficiency is of the honeycomb pattern. The yieldmap shows that it has a 100% active surface area. However, the hexagon pattern does not match thesquare pitch of the pixel chip. Using a planar anode, it can be potentially 100%. yield map showedthat the metamaterial has a 100% active surface with higher transmission yield near the verticalwalls of the hollow ribs. Depending on the read-out, this means a collection efficiency of 100%– 12 – a) (b)
Figure 13 : Yield map of octagon tynode with ribs facing downwards. The reflection yield mapmatches the SEM image. The former is determined from the sample current, while the latter isdetermined from SE emission from the sample. The glow near the edge of the octagon/square isvisible on both pictures. The transmission yield near the vertical walls is higher due to the angle ofincidence of the PEs.when for instance a planar anode is used. For TiPC, the collection efficiency would be definedas the number of photoelectrons emitted from the photocathode can trigger a signal in a pixel. Inthis case, the focusing of the octa pattern can direct most electrons to the pixel pad when they areemitted from the octagon. Using the pattern presented here, the collection efficiency would be thesurface of the octagon divided by a pixel surface. The collection efficiency is then just 68%. If weassume the SEs generated within the ribs ends contributes as well, then it is 82%. The remaining18% are the squares in the pattern, which will form a channel on their own which ends next to thepixel pads. A solution is to use a different pattern for the first tynode. For instance, a simple squarepattern. This pattern is less strong. A solution to guarantee that electrons in the ribs are collectedas well is to use a different shape, such as a U- or V-shaped beams. This can be achieved by usingdifferent recipes in DRIE or using wet chemical etching using KOH, which is anisotropic.The largest metamaterial membrane that were made has a surface of approx. 12 . . This– 13 – a) (b) Figure 14 : Yield map of octagon tynode with ribs facing upwards. The transmission yield map ismore homogeneous. In reflection, the vertical walls of the ribs generate a lot more Reflection SEsdue to the angle of incidence of the PEs.is much larger than achievable with a flat membrane. Though, this is almost at the limit. The filmsare prone to breaking due to electrostatic attraction, deformation by pressure differences and, inone case, by simply looking at it. The window surface is designed to cover 64x64 pixels. One cansacrifice a few rows of pixels to create a silicon frame with smaller windows. For instance, 30x30pixels windows and a frame with 4 pixels width (0 .
22 mm)
Mechanical metamaterials can be functionalized as transmission dynodes by using a tri-layerAl2O3/TiN/Al2O3 film deposited by atomic-layer-deposition. The 3D structure of the metama-terial allows films to have an extreme surface/thickness ratio in comparison with flat membrane.Meta-material can have a 100% active surface over a large-surface. The largest surface area wasapproximately 12.4 mm2 with a thickness of 10/5/15 nm for a tri-layer Al O /TiN/Al O film. Thetransmission yield response of the metamaterial with is measured by determining the yield map.Although there are some variations in transmission yield across the surface, transmission secondary– 14 –lectron emission is measured. The metamaterial has been successfully functionalized by using atri-layer film. In addition, the hexagon/octagon cells can be operated in focusing mode by facing thecell towards the next target cell. Facing upside down gives a more uniform response. (COMSOL).The thickness needs to be reduced to improve the transmission electron emission performance.This can be achieved by either improving the final release steps by improving the selectivity of theDRIE recipe towards Si and Al2O3. Or, a different road is to create a silicon frame in which thethinner metamaterial is suspended. The open surface would be intersected by a few silicon beams.By sacrificing a few rows of pixels, the silicon frame can provide additional strength. The collectionefficiency of the tynodes can be further improved by using U- or V-shaped ribs. SEs generatedwithin a rib would then be extracted in either side of the rib. Other patterns can be designed for thefirst tynode, which can collect more PEs towards a tynode âĂŸchannelâĂŹ. And furthermore, theTSEE of the metamaterial can be increased by using different ALD materials, such as MgO. References [1] 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.[2] T. Hakamata, Ed.,
Photomultiplier tubes , 3rd ed. Hamamatsu Photonics K.K., 2007.[3] S. X. Tao, H. W. Chan, and H. Van Der Graaf, “Secondary electron emission materials for transmissiondynodes in novel photomultipliers: A review,”
Materials , vol. 9, no. 12, 2016.[4] M. Hagino, S. Yoshizaki, M. Kinoshita, and R. Nishida, “Caesium Activated CsI Transmission-typeSecondary Emission Dynode,”
Advances in Electronics and Electron Physics , 1972.[5] 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.[6] H. Chan, V. Prodanović, A. Theulings, C. Hagen, P. Sarro, and H. v.d. Graaf, “Secondary ElectronEmission from Multi-layered TiN/Al2O3 Transmission dynodes,”
Arxiv preprint , vol. -, no. -, pp. –,2020.[7] V. Prodanović, “Ultra-thin mems fabricated tynodes for electron multiplication,” Ph.D. dissertation,Delft University of Technology, 2019.[8] K. Davami, L. Zhao, E. Lu, J. Cortes, C. Lin, D. E. Lilley, P. K. Purohit, and I. Bargatin, “Ultralightshape-recovering plate mechanical metamaterials,”
Nature Communications , vol. 6, p. 10019, 2015.
A Secondary electron yield map