Magnetic patterning of (Ga,Mn)As by hydrogen passivation
Laura Thevenard, Audrey Miard, Laurent Vila, Giancarlo Faini, Aristide Lemaître, Nicolas Vernier, Jacques Ferré, Stéphane Fusil
aa r X i v : . [ c ond - m a t . m t r l - s c i ] J un Magneti patterning of (Ga,Mn)As by hydrogen passivationL. Thevenard, ∗ A. Miard, L. Vila, G. Faini, and A. LemaîtreLaboratoire de Photonique et de Nanostru tures CNRS, Route de Nozay, 91460 Mar oussis, Fran eN. Vernier and J. FerréLaboratoire de Physique des Solides CNRS (UMR 8502), Université Paris-Sud, 91405 Orsay, Fran eS. FusilUnité Mixte de Physique CNRS-Thalès (UMR 137) RD 128, 91767 Palaiseau Cedex, Fran e(Dated: 3 novembre 2018)We present an original method to magneti ally pattern thin layers of (Ga,Mn)As. It relies onlo al hydrogen passivation to signi(cid:28) antly lower the hole density, and thereby lo ally suppress the arrier-mediated ferromagneti phase. The sample surfa e is thus maintained ontinuous, and theminimal stru ture size is of about 200 nm. In mi ron-sized ferromagneti dots fabri ated by hydrogenpassivation on perpendi ularly magnetized layers, the swit hing (cid:28)elds an be maintained loser tothe ontinuous (cid:28)lm oer ivity, ompared to dots made by usual dry et h te hniques.PACS numbers:A tive resear h on the diluted magneti semi ondu tor(DMS) Ga − x Mn x As is now qui kly at hing up with theextensive knowledge a umulated on metalli ferroma-gnets. However, envisioning future appli ations for thismaterial presents several hallenges. In addition to rai-sing the Curie temperature [1℄, it is ne essary to unders-tand the me hanisms of the magnetization reversal insingle-domain systems. Novel magneti behaviors are in-deed expe ted to arise, where demagnetizing, edge, and(cid:28)nite volume e(cid:27)e ts ome into play, as the dimensionsbe ome omparable to the typi al magneti lengths (do-main width, domain wall (DW) width, and ex hangelength).Fabri ation of DMS mi rostru tures is usually doneusing dry et h te hniques [2, 3, 4, 5℄. While having provi-ded experimental results on DW velo ity [3, 4℄, this ap-proa h may however reate an edge roughness along thinpatterns. These irregularities are likely to grip the DW,indu ing high depinning (cid:28)elds and pro ess-dependentmeasurements.In this paper, we present a novel magneti patterningte hnique, designed to ir umvent this in onvenien e. Ta-king advantage of the arrier-indu ed ferromagnetism inGa − x Mn x As, we use atomi hydrogen to form ele tri- ally ina tive omplexes with the magneti impurities,thereby suppressing the ferromagneti phase [6, 7, 8℄. Bydepositing a mask before the passivation, we pattern aperpendi ularly magnetized layer into mi ron-sized ferro-magneti dots. After the hydrogenation pro ess, the maskis removed. Only the zones shielded from the plasma re-main ferromagneti , while the rest of the layer is para-magneti . Interestingly, this patterning route is rever-sible, sin e a low temperature anneal an restore theproperties of the initial, non-patterned layer, by brea- ∗ Ele troni address: laura.thevenardlpn. nrs.fr king the (Mn,H) omplexes [9, 10℄. The main advantagesof this purely di(cid:27)usive pro ess are to pattern reversiblya (Ga,Mn)As layer, ontrary to ion implantation or et- hing, and to maintain the ontinuity of the (cid:28)lm, whi hallows near (cid:28)eld mi ros opy investigations and smoo-thing of border e(cid:27)e ts. Note that this patterning pro esshad been proposed in earlier ommuni ations [11, 12, 13℄.The sample was grown by Mole ular Beam Epitaxyfollowing the pro edure des ribed in detail in Ref. [14℄.It onsists of a 50 nm Ga . Mn . As layer grown on arelaxed Ga . In . As bu(cid:27)er, with a Curie temperatureof 118 K after annealing and an e(cid:27)e tive perpendi ularanisotropy (cid:28)eld of H u ≈ Oe at 2 K. In a previouspaper [14℄, we showed on a very similar (cid:28)lm that themagneti easy axis was along [001℄ at all temperatures.Rare, growth-indu ed defe ts prevented the periodi stri-ped arrangement of the domain stru ture expe ted fora perfe t uniaxial ferromagneti (cid:28)lm. However, domainswere homogeneous and 20 µ m wide at 80 K, making themsuitable for patterning single-domain magneti elements.The patterning was done in the following way : after areful desoxydation, a titanium mask was (cid:28)rst depositedby e-beam lithography lift-o(cid:27). A 40 nm thi kness was suf-(cid:28) ient to prevent the hydrogen atoms from entering thelayer. The mask onsisted of three 200 × µ m arraysof dots, with sizes (and spa ings) : 10 µ m (10 µ m), 5 µ m(10 µ m) and 1 µ m (5 µ m). The sample was then exposedto a hydrogen plasma during 2 h ; it will be referred toin the rest of the paper as "sample H ". Transport mea-surements showed an in rease of the sheet resistivity bytwo orders of magnitude, on(cid:28)rming that the layer hadindeed been passivated. For omparison, another set ofarrays, "sample E ", was then pro essed by et hing, usingthe same layer and mask. Ion Beam Et hing (IBE) at a20 ◦ in iden e angle was used to ensure 80 nm high, verti- al sidewalls. Finally, the metalli mask was removed o(cid:27)both samples by a diluted HF solution. Here, we empha-size that the surfa e of sample H was then ompletely ontinuous, whereas sample E was left stru tured.Fig. 1: (a)-( ) : Kerr mi ros opy snapshots (T=85 K) du-ring the magnetization reversal of dots patterned by hydro-gen passivation. A 5 s (cid:28)eld pulse of H=100, 100, and 120 Oewas applied to dot arrays of 10, 5 and 1 µm . The ontrast islow for the latter onsidering the resolution of our PMOKEmi ros ope ( λ = 1 µm ). (d) Gaussian (cid:28)t of the numberof reversed 10 µm dots, to extra t a mean swit hing (cid:28)eld H sw = 243 ± Oe and a (cid:28)eld dispersion σ = 26 ± Oe (T=20K).The magnetization reversal was then investigated bet-ween 4 K and 100 K by polar magneto-opti al Kerr(PMOKE) mi ros opy, using a set-up des ribed in Refs.[15, 16℄. This te hnique yields an ex ellent ontrast forGa − x Mn x As thin (cid:28)lms [14℄ and is parti ularly well sui-ted to the study of our samples sin e the signal is sensitiveto the perpendi ular magnetization omponent. Snap-shots were taken in zero (cid:28)eld after applying s pulsesof positive magneti (cid:28)eld, and substra ted from a refe-ren e image taken at negative saturation, to highlightthe magneti ontrast between up- and down- magneti-zed areas.Field pulses of in reasing amplitude were applied until omplete magnetization reversal of the array. Images ta-ken on sample H during the reversal (Fig. 1a- ) learlyshow that the passivation te hnique was su essful in pat-terning magneti ally the layer into 1, 5 and 10 µm stru -tures. The arrays of et hed dots follow a omparable ma-gneti behavior (not shown here). As expe ted from thelow saturation magnetization, we moreover veri(cid:28)ed thatdipolar or ex hange intera tions between dots played aminimal role, sin e the number of reversed dots followedexa tly an independent-dot probability law [17℄.By plotting the number of dots that swit hed betweentwo onse utive (cid:28)eld pulses and (cid:28)tting it to a gaussian,we then extra ted a mean swit hing (cid:28)eld, H sw , and the orresponding (cid:28)eld dispersion σ for a given dot array andtemperature (Fig. 1d). H sw values obtained by this pro- edure were reprodu ible within ±
20 Oe. The resultingswit hing (cid:28)elds for 5 and 10 µm dot arrays on samples H and E are shown in Fig. 2, along with the oer ive (cid:28)eld, H c , of the non-patterned layer determined by PMOKE.Fig. 2: Mean swit hing (cid:28)elds of 10 µm (full symbols) and 5 µm (empty symbols) dots patterned by hydrogenation (bla ksquares) and by et hing (red ir les). The oer ive (cid:28)eld H c of the non-patterned layer was also measured by Kerr e(cid:27)e t.Above 100 K, the ontrast was too weak to provide reliablemeasurements.The mean swit hing (cid:28)elds of sample E are systemati- ally mu h higher than H c , and de rease by about when the perimeter of the dot doubles. H sw varies mo-notonously from 550 to 200 Oe when the temperaturein reases from 4 K to 100 K, with a fairly large relativedispersion σH sw from 22 % at 4 K to 30 % at 100 K. Dotarrays of sample H have a quite di(cid:27)erent behavior. Swit- hing (cid:28)elds are at least twi e smaller than those of sample E , while remaining above H c . They are also better de(cid:28)-ned at all temperatures, with a relative dispersion σH sw from 10 (T=4 K) to 18 % (T=100 K). Most remarkableis the non-monotonous evolution of H sw with tempera-ture, rea hing a maximum at T max =40 K (resp. 60 K) for10 % µm (resp. 5 µm ) dots. This unusual behavior wasreprodu ible after temperature y ling.In both samples, no partly-swit hed dot was observedin the a essible (cid:28)eld and time range. The swit hing (cid:28)eldsare however (cid:28)ve to ten times smaller than those expe -ted for a single-domain, oherent reversal ( H sw = H u ),suggesting a nu leation/propagation reversal me hanism.On the non-patterned layer, the magnetization reversalwas shown to be triggered at rare nu leation enters (1 mm − ), before rapidly developing by easy domain-wallpropagation [14℄. Let us (cid:28)rst onsider the et hed dots,sample E . Upon redu ing the size of the system, the nu- leation be omes more di(cid:30) ult sin e the probability of(cid:28)nding nu leating enters with low energy barriers is de- reased by a simple geometri al e(cid:27)e t [17℄. From the weaksensitivity of H sw on the area of et hed dots, it an be in-ferred that another me hanism is ertainly involved, su has nu leations initiated on the rough edge of the dot, ashas already been observed on metalli mi rostru tures[18℄. The perimeter redu tion of the dot entails a de rea-sed probabibility of (cid:28)nding a nu leating defe t along theedge, and therefore leads to a higher H sw . The thermala tivation of the nu leation pro ess is responsible for thede rease of swit hing (cid:28)elds with temperature. After nu- leation, the reversal pro eeds by fast propagation withinthe dot.Another me hanism has to be invoked for sample H ,similar to that eviden ed in Pt/Co/Pt dots patternedby fo used Ga ion beam irradiation [19℄. The hydro-gen passivation over the mask results in a hole densitygradient within a ring around the ferromagneti dot, agradual interfa e between the dot and the paramagneti (Ga,Mn)As :H matrix. This soft magneti zone is likely tobe an easy nu leation region from whi h the magnetiza-tion reversal is initiated. The dispersion is also expe tedto be mu h thinner than in sample E , sin e the reversalby wall propagation inside the dot no longer depends ona statisti al distribution of defe ts around the dot edge,but on a fairly smooth, soft magneti interfa e, hen e re-sulting in the lower values of H sw and σH sw for sample H .In the soft ring around the dot, the hole density ishigh enough to yield a ferromagneti phase, but su(cid:30)- iently low to indu e an in-plane magnetization, as hasbeen predi ted by mean-(cid:28)eld theories [20℄, and shownexperimentally [9, 21℄. This may be ompatible with theunusual temperature dependen e of H sw , sin e the in-plane magnetization at low temperature and low holedensity is expe ted to (cid:29)ip out of the plane, along [001℄,with in reasing temperature [10, 21℄. The swit hing (cid:28)eldswill then in rease sin e spins will be blo ked up to the oer ive (cid:28)eld, instead of rotating ontinuously with theapplied (cid:28)eld, along a hard-axis-like urve. The behaviorof H sw may therefore re(cid:29)e t the temperature evolutionof the magneti anisotropy in the ring around the ferro-magneti dot.In order to estimate the width of the hydrogen di(cid:27)u-sion front that onstitutes a soft magneti ring aroundthe dots, we used ondu tive-tip Atomi For e Mi ro-s ope (CT-AFM) measurements to establish a resistivitymapping of the dot. In (cid:28)rst approximation, we assumedit to be proportional to the hydrogen density. This room-temperature te hnique asso iates a standard AFM witha voltmeter onne ted between the p -doped diamond tip(sample surfa e) and a ba k onta t. Thus, it is prefe-rable to work with ondu tive substrates. Ferromagneti 4 µm wide dots were therefore fabri ated following thesame te hnologi al pro ess, on a 50 nm Ga . Mn . Aslayer grown over GaAs :Be (Fig. 3a).After removing the metal mask, a (cid:28)rst regular AFMs an of the dot showed a smooth surfa e with a 2 nmroughness (Fig. 3b). The CT-AFM measurement theneviden ed an ele tri al pattern on the layer identi al tothe shape of the mask (Fig. 3 ). The pro(cid:28)le taken alongthis s an on(cid:28)rmed that the resistivity in reased by twoorders of magnitude outside the dot (Fig. 3e). The inter-fa e pro(cid:28)les were then adjusted by a standard di(cid:27)usionerror fun tion. This is a fairly rude approximation sin eit entails a 1-D di(cid:27)usion pro ess [22℄. However, averagedover a dozen pro(cid:28)les, it yielded a hara teristi di(cid:27)usion Fig. 3: (a) AFM image of the mask used for the patterning.Note that its irregular shape is a non-intended lithographyartefa t. (b) AFM image after mask removal : the surfa eis ontinuous. ( ) Condu tive-Tip AFM s an after mask re-moval, revealing an ele tri al pattern exa tly identi al to themask's. (d,e) The logarithmi pro(cid:28)le eviden es a two ordersof magnitude de rease of the resistivity over about 80 nm. Inthis ase, the hara teristi di(cid:27)usion length was estimated byan error fun tion (cid:28)t (line) to about 20 nm.length of 27 nm, and a mean di(cid:27)usion front width of87 nm (Fig. 3d), a value ompatible with Kerr mi ro-s opy experiments done on small stru tures. This dis-tan e being larger than the DW width ( ≈
20 nm [23℄)and the ex hange length ( ≈≈