56 CHAPTER 3. SELF-ORGANIZED NPS ARRAYS: MORPH. ASPECTSfig. 3.5(c) after the deposition <strong>of</strong> ≈ 5 nm <strong>of</strong> gold. The surface is still characterizedby coherently aligned elongated structures, indicating that the overgrown layer has followedthe structure <strong>of</strong> the underlying template. The ordering is also confirmed by the2D Fourier spectrum, where the central peak is still quite stretched in the [0¯11] direction;however, a secondary background is also present, slowly decaying from the central peakin all directions uniformly. This indicates that the metallic stripes are not completelycontinuous but rather composed <strong>of</strong> close-packed grains; it also suggests that the growth <strong>of</strong>gold on LiF does not likely proceed layer-by-layer but probably follows a layer-plus-isl<strong>and</strong>(Stranski-Krastanov) mode. Due to the finite size <strong>of</strong> the AFM tip, it is not possible todeduce from the AFM images whether the isl<strong>and</strong>s coalesced or remained separated. Nevertheless,the lateral size <strong>of</strong> the grains seems to be limited to an upper size <strong>of</strong> ≈ 15 nm.This is shown in fig. 3.7(c,d), where we report AFM images measured on two sampleswith different ripple spacing <strong>and</strong> the same amount <strong>of</strong> deposited gold, ≈ 5 nm; in panel(c), the sample has a periodicity Λ ≈ 25 nm <strong>and</strong> linear chains <strong>of</strong> grains are observed,where the grains lateral sizes matches the grooves width; on the contrary, in panels (d)<strong>and</strong> (b), for a spacing Λ > 40 nm, no isl<strong>and</strong> as large as the ripples is found, <strong>and</strong> insteadmultiple chains are formed inside each groove.a.b.[001] [001]c.[001]d.[001]Figure3.8: AFMimages<strong>of</strong>goldnanoparticles<strong>arrays</strong>onnanopatternedLiF(110), followingthe deposition <strong>of</strong> t Au gold <strong>and</strong> annealing at T ≈ 100 ◦ C. Panel a: Λ = 30 nm, t Au =4.5 nm. Panel b: Λ = 40 nm, t Au = 4.8 nm. Panel c: Λ = 45 nm, t Au = 5.4 nm. Paneld: Λ = 55 nm, t Au = 7.2 nm.Given the natural tendency <strong>of</strong> gold to form agglomerates, we can further promotethe dewetting <strong>of</strong> the “nanowires” (NW) by mild annealing the substrates [189–191]. In
3.2. 2D ARRAYS OF GOLD NANOPARTICLES 57fig. 3.8(c) we report an AFM image measured on the same sample <strong>of</strong> fig. 3.7(b) after theannealingat400 ◦ C.Comparingthetwoimages, themaineffect<strong>of</strong>theannealingprocedurewas a partial melting <strong>of</strong> the gold grains, which merged together forming larger particles,with a size comparable to the ripples width. The underlying ripple structure assisted theprocess, preventing the spreading <strong>of</strong> the grains across the ridges, <strong>and</strong> instead promotingthe formation <strong>of</strong> parallel chains <strong>of</strong> nanoparticles, preserving the same periodicity Λ <strong>of</strong> theripples.Interestingly, the LiF grooves not only guided the particles arrangement, but alsoinfluenced their sizes <strong>and</strong> shape. This effect can be observed looking at fig. 3.8, wherewe reported AFM images <strong>of</strong> several NP <strong>arrays</strong>, annealed at the same temperature <strong>of</strong>400 ◦ C but fabricated under different growth conditions. Let’s first consider panels (a)<strong>and</strong> (b). The samples have, respectively, periodicities Λ ≈ 30 nm <strong>and</strong> ≈ 40 nm, <strong>and</strong>a similar amount <strong>of</strong> gold were deposited, ≈ 4.5 nm <strong>and</strong> ≈ 4.8 nm. In both cases theparticles have an in-plane circular shape, <strong>and</strong>, although it is not possible to preciselydetermine the particles dimensions due to the convolution <strong>of</strong> the AFM tip, their typicalsize increases according to the variations <strong>of</strong> width <strong>of</strong> the LiF grooves. In panel (c) a samplewith Λ ≈ 45 nm, similar to the sample in panel (b), but t Au ≈ 5.4 nm <strong>of</strong> deposited goldis shown. In this case, the particles are still confined inside the ripples, <strong>and</strong> the higheramount <strong>of</strong> gold led to the formation <strong>of</strong> more elongated structures, with an average aspectratio <strong>of</strong> ≈ 1.5. A similar trend is observed increasing further the thickness <strong>of</strong> gold. Thesample in panel (d) has Λ ≈ 55 nm <strong>and</strong> t Au ≈ 7.2 nm.The majority <strong>of</strong> the particles is again as wide as the grooves, however we also observean increased dispersion <strong>of</strong> particles sizes with respect to the previous cases, both along<strong>and</strong> across the ripples. This is probably due to the presence <strong>of</strong> morphological defects inthe underlying ripple structure, <strong>and</strong> it happens in concordance with the observations <strong>of</strong>the previous section, where an increasing disorder <strong>of</strong> the surface morphology was foundat the largest Λ (fig. 3.5(d)). Looking at the shape, the particles now have a maximumaspect ratio <strong>of</strong> ≈ 2, higher than for sample (c), but many less anisotropic ones are alsovisible; this could be due to the fact that the formation <strong>of</strong> structures with aspect ratiosgreater than 2 is not favoured, so in these cases a segregation to smaller particles occurs.Then, we can conclude that, for periodicities between 25 nm <strong>and</strong> 60 nm <strong>and</strong> thickness <strong>of</strong>deposited gold between 4 nm <strong>and</strong> 10 nm, the particles width is fixed by the grooves widthwhile the shape (aspect ratio) is determined by the amount <strong>of</strong> deposited gold.We can interpret this effects by considering the dynamics <strong>of</strong> the Au nanowires dewetting.If the NW evolve following a Rayleigh-type instability [192], then a linear relationshipbetween the NP spacing (<strong>and</strong> hence their volume) <strong>and</strong> the initial NW radiusr (r ∝ (Λ · t Au ) 1/2 ) is expected [193], showing that both the mean spacings along theripples <strong>and</strong> volume <strong>of</strong> the NP get larger in correspondence <strong>of</strong> the increasing Λ <strong>and</strong> t Au .The simultaneous increase in the aspect ratio can be ascribed to the lateral “constraint”effect exerted by the LiF nanopatterns, that triggers an anisotropic NP growth during thedewetting process.Another unexpected feature <strong>of</strong> the NP <strong>arrays</strong> is the relative positions <strong>of</strong> particlesbelonging to different grooves. As the dewetting induced by the annealing takes placeseparately in each ripple, one may expect no transversal correlation between the position<strong>of</strong> the particles in adjacent rows; in contrast, many regions can be distinguished wherethe particles are locally arranged on a rectangular grid. These regions can be emphasizedby looking at the autocorrelations <strong>of</strong> the AFM images. In fig. 3.9 we report the 2Dautocorrelation plots for the samples <strong>of</strong> fig. 3.8(b,c); both the figures are characterizedby rectangular patterns <strong>of</strong> peaks <strong>and</strong> hollows, with sides oriented in the [1¯10] <strong>and</strong> [001]