WO2000001866A1

WO2000001866A1 - Method for crystallising a semiconductor material and crystallising system

Info

Publication number: WO2000001866A1
Application number: PCT/FR1999/001598
Authority: WO
Inventors: Pierre Legagneux; Christian Collet; Didier Pribat
Original assignee: Thomson-Csf
Priority date: 1998-07-03
Filing date: 1999-07-02
Publication date: 2000-01-13
Also published as: FR2780736B1; FR2780736A1

Abstract

The invention concerns a method for crystallising a semiconductor material layer (1) using an energy beam (F1) illuminating on one surface (L x l) the semiconductor material layer (1). The layer is crystallised by the relative movement of the beam at the layer surface along a specific direction (OX). The beam has an energy profile along the displacement direction (OX) such that there is at least a first energy level for obtaining coarse grain crystal lisation (Super Lateral Grow-SLG) from a small grain amorphous or crystalline phase with of the semiconductor material and one or several second energy levels higher than the first energy level and located towards the front of the zone illuminated by the beam along the beam displacement direction (OX). Said second energy level(s) enable to melt/dissolve a material comprising both coarse grain and small grain crystalline phases. The invention is useful for producing coarse grain crystalline semiconductor material layers.

Description

METHOD OF CRYSTALLIZATION OF A SEMICONDUCTOR MATERIAL AND CRYSTALLIZATION SYSTEM

The invention relates to a method for crystallizing a layer of semiconductor material and to a system for carrying out this method.

More particularly, it involves producing semiconductor layers having a uniform coarse-grained microstructure for the manufacture of thin-film transistors with improved performance for large-area liquid crystal displays with an active polysilicon matrix, for example.

The market for liquid crystal devices is growing rapidly (between 10 and 20% in recent years) and could double by the year 2000. A pioneering technology based on amorphous silicon (aSi), we have successively seen l introduction of polycrystalline silicon (pSi):

1) in a high temperature pSi technology> 600 ° C on quartz substrate, limited in size; 2) in a low temperature <500 ° C pSi approach allowing the introduction of glass substrates at low cost and large surface area. Only this last approach seems able to overcome the new obstacles - that the market seems to want to impose with devices at 2000 * 1600 pixels in steps of 200 microns or less at the edge of the year 2005. We can consider screens of '' a size up to 40 inches in 16: 9 made on glass plates which a committee of 30 major manufacturers in the field defined the size at 960mm ^* 1100mm. Only the use of pSi with mobilities of electrons and holes of 2 orders of magnitude greater than those of aSi could respond to this densification of integration with peripheral control circuits directly integrated on the screen, rate increases aperture and screen size.

The goal is therefore to obtain the most uniform layers of pSi possible, composed of the largest grains, without internal defects. Technological mastery of a large grain size is one of the keys to obtaining high mobility for load carriers. The intuitive approach to direct chemical deposition of pSi in the vapor phase (CVD) around 600 ° C [D. Pribat and al "Science and Technology of Thin Film", FC Mattacotta and G. Ottaviani Eds. (Word Scientific, Singapore 1995) p 293.359 (1995)] does not give satisfactory results due to a high surface roughness disturbing high mobility. We therefore have recourse to a crystallization of an amorphous precursor.

This crystallization can be carried out in the solid phase by means of an oven, around 600 ° C., for several hours or even tens of hours. The grains obtained can be large (up to 5 microns) but unfortunately filled with intra-granular defects, cause of strong electronic degradation of the material and consequently of the devices produced.

In addition, for an industrial approach this step is highly costly in terms of production time. As a consequence, a crystallization path using pulsed excimer laser [T. Sameshima MRS Symp Proc 71 (1986), 435] has been studied and seems to currently favor industrialists.

The short pulses (a few tens of nanoseconds) create weak thermal peaks with little or no damage to the substrates. The working frequencies up to 300 Hz to date, reasonable for an industrial approach, make this type of laser very attractive.

A single impulse (single shot) approach seems a priori very

2 elegant. Tests were carried out on 3 * 3 cm surfaces. Seen the

-2 fluences necessary for crystallization (300 to 500 mJ.cm) and the increase in the size of the screens, these high-power single-shot machines do not seem able to process a sample in a single pass, so they lose a good part of their interest and are prohibitively expensive. We are therefore moving towards multishot irradiation (that is to say scanning the surface) with recovery rates by a light brush with a width less than 1 mm and as long as possible (from a few cm to a few tens of cms), ideally of a length equal to the smallest of the dimensions of the glass plates used to avoid a second lateral pass and an additional overlap. The family of excimer lasers is grouped on 5 different wavelengths depending on the gases used: F2 (157 nm), ArF (193 nm), KrF (248 nm), XeCI (308 nm) and XeF (351 nm).

FIG. 1 schematically represents a curve giving the grain sizes of crystallized polysilicon samples as a function of the energy density of laser irradiation of the sample.

We see that gradually increasing the energy density, first we obtain a material of small grains with more or less constant size slightly increasing over a large energy range (partial melting of the material more and more deep) then a sudden increase in grain size in a very narrow energy window, the so-called “SLG” (Super Lateral Grow) zone, total melting limit of the thickness of the material. This SLG regime corresponds to the reminiscence of a few germs disseminated at the interface, from which the solidification of the liquid phase will take place. When the energy increases further, these residual germs disappear by fusion and one obtains a polycrystal with very small grains, unusable in term of electronic transport (all the molten layer) (see the article "Crystallization of amorphous silicon by laser with excimers of D. PRIBAT et al published in the Annales de Physique, C1-213, vol. 22, February / April 1997).

The problem is to obtain a polycrystalline material with uniform grains. To do this, we generally use a laser beam whose energy distribution is very flat (“top-hat” beam) (see for example the recent article by Kiyoshi YONEDA, in “Conférence record of the 1997 International Display Research Conférence” , published by the "Society for Information Display", 1526 Brookhollow Drive, Santa A, CA 92705-5421, page M-40).

The disadvantage comes from a repetitive marking due to the energy slope of the front before the impulse. The irradiated area under this front face will have a mixed and complex structure (mixture of small and large grains). After advancing the laser spot, the maximum energy of the laser in its flat part will be insufficient to completely recast this area, hence locally the presence of a different crystalline structure generating an indelible marking by this method of crystallization and this energy profile. beam. This drawback will be better understood on reading Figures 2a to 2e and their commentary.

Indeed, the pulses used have an energy profile as shown in Figures 2a and 2b. FIG. 2a represents a layer 1 of amorphous semiconductor material illuminated by a beam F1 supplied by a light source S1. The surface illuminated by the beam has the dimensions I x L. FIG. 2b represents the energy profile of the pulse measured in the direction OX (according to dimension I). Its higher energy level corresponds to the fusion of the amorphous semiconductor Ta.

Figure 2c shows in section the semiconductor layer 1 and highlights the action of the light pulse on the semiconductor material. The pulse illuminates an area Z1. The document by D. PRIBAT et al published in “Annales de Physique” C1-213, vol. 22, February / April 1997 describes the problem of crystallization of an amorphous semiconductor material under the action of such a pulse.

In the SLG1 zone, the originally amorphous material is melted exactly over the entire thickness and crystallization gives rise to the formation of a homogeneous crystallized material with large grains. In zones z1 and z2, after fusion, crystallization gives rise to a crystallized material containing both large grains, and small grains due to the fact that in the front and rear flanks of the pulse, we are in the presence of an energy gradient. These zones are therefore of mediocre and heterogeneous crystalline qualities. After this last pulse, if the light beam is displaced along the axis OX and a surface which overlaps the previous illuminated surface is illuminated, we have a configuration shown in FIGS. 2d and 2e. The light pulse of Figure 2d is offset from the pulse of Figure 2b. It lights up a zone Z'1. However, the energy of the pulse is insufficient to remelt the already crystallized material up to the interface because the latent heat of transformation of the polycrystal is higher than that of the amorphous. The zone Z1 therefore remains practically intact and in particular the zone z2 of different microstructure. Under these conditions, only the SLG'1 zone gives rise to a coarse-grained crystallization (see FIG. 2e). Zone z'2 becomes similar to zone z2. So we see that after treatment by a succession of light pulses displaced on the surface of the layer of semiconductor material, this will present an alternation of zones such as SLG1 and SLG'1 made of coarse-grain semiconductor material and zones such as z2 and z'2 made of heterogeneous coarse / small grain material. This heterogeneity of microstructure is extremely detrimental to the uniformity of the characteristics of the transistors which will then be produced in these layers. A periodic marking, the origin of which has been previously explained, may be superimposed on a random marking due to the instabilities from one laser to the next (temporal instability). The spatial stability of excimer lasers in the current state of the art (3rd generation of laser) at ± 5% is insufficient to avoid this type of marking in an industrial approach, large irradiated area (K. YONEDA, State of the art temperature processed Poly-Si TFT Technology, SID Toronto 09/1997). Even reduced to ± 2% by optical homogenization systems (see K. YONEDA already cited), these energy differences result in random grain size variations detrimental to the regularity of the electronic properties of components made from this material . The invention therefore relates to a method for obtaining a uniformly coarse-grained semiconductor layer.

The invention therefore relates to a method of crystallization of a layer of semiconductor material using an energy beam illuminating on a surface (L x I) the layer of semiconductor material, the crystallization of the layer being effected by displacement of the beam on the surface of the layer in a determined direction, characterized in that said beam has an energy profile in the direction of movement such that there is at least a first energy level allowing a crystallization to be obtained at large grains (Super Lateral Grow type - SLG) from an amorphous or crystalline phase with small grains of the semiconductor material as well as one or more second energy levels higher than the first energy level and located towards the front of the area illuminated by the beam in the direction of movement of the beam, and allowing a fusion / dissolution of a material comprising both large-grain crystalline phases and s small grain crystalline phases. The invention also relates to a system for crystallizing a layer of semiconductor material comprising a light source illuminating, using a light beam, a face of the layer of semiconductor material as well as means for moving in one direction and a direction determined the light beam with respect to the layer of semiconductor material characterized in that it comprises means between the source and the layer of semiconductor material for inducing an energy gradient of the beam.

The various objects and characteristics will appear more clearly in the description which follows, given by way of example and in the appended figures which represent:

- Figures 1 to 2e, techniques of the known art and described above;

- Figures 3a to 3h, the production method according to the invention; - Figures 4a and 4b, a variant of the process of Figures 4a to

4h;

- Figures 5a to 5f, an operation of the method of the invention with unstable pulses;

- Figures 6a to 6c, alternative energy profiles of a light pulse used in the context of the invention;

- Figure 7, a system implementing the invention;

- Figure 8, a comparison of the operation of the system of the invention with pulses of different energy gradients - Figures 9 and 10 show an alternative embodiment of the invention for obtaining parallel crystallized bands.

To eliminate, by total fusion, the defective crystallized zones such as z2 and z'2 of FIG. 2e, provision is made to illuminate the layer of semiconductor material using a beam having an energy gradient according to the direction beam OX displacement.

Each light pulse therefore has an energy profile as shown in FIG. 3a. It comprises at least one energy level allowing a fusion / dissolution of a crystalline material to transform it into homogeneous crystalline material with small grains. In Figure 3a, these are energy levels between e3 and e4. It also includes at least one energy level allowing the melting / dissolution of a crystalline material with small grains (dimension 50 nm for example) to transform it into crystalline material with large grain (dimensions 1 μm for example). In Figure 3a, these are the energy levels between e2 and θ3.

In FIGS. 3a to 3h, pulses having a linear energy variation according to the direction of movement of the beam with respect to the layer of semiconductor material have been used. However, the pulses could have any other form, such as for example that shown in FIG. 6a, provided that it has an energy level between e3 and e4 and an energy level between e2 and e3.

Referring to Figures 3a to 3h, we will now describe the operation of the method of the invention. Assume for example that the layer of semiconductor material 1 is amorphous silicon, previously deposited by a method known to those skilled in the art and containing a residual hydrogen level compatible with laser irradiation (typically of the order 1 to 2%). The light pulse whose energy profile has the form shown in FIG. 3a is transmitted to layer 1.

The range of energy levels e0-e1 makes it possible to transform the amorphous silicon of layer 1 into a large grain polycrystal. This is represented by the area a2 in FIG. 3b; this figure shows in section the structure of the material obtained after irradiation. In other words, this energy range makes it possible to melt exactly and up to the interface, layer 1. In this particular case, the solidification of the liquid takes place from crystalline seeds remaining at the liquid interface. / substrate (these crystalline seeds were synthesized beforehand by explosive crystallization, as described in the article already quoted from D. PRIBAT et al.) and one is in the growth regime of the “SLG” type, which has been previously described. Typically and without these values being limiting, the energy range e0-e1 is located around 400-415 mJ / cm ² , for a semiconductor layer thickness of the order of 80 to 100 nm. The slope of the energy profile is for example 20%, but this is not limiting. Energy levels higher than e1 give rise to the complete fusion of layer 1, with disappearance of the crystalline germs at the interface. The solidification of the liquid phase is now carried out by homogeneous nucleation within this same liquid and the material obtained after crystallization has a microstructure with small grains (see FIG. 1). This is schematically represented by the area a3 in FIG. 3b. Finally, it can also be seen in this same FIG. 3b that the zones corresponding to the sides of the energy profile give rise to a mixed structure, composed of small and large grains (zones ai and a4), phenomenon which has been explained previously.

In FIGS. 3c and 3d, the energy beam has been translated (by means not shown) on the surface of the semiconductor layer, in the direction OX, in such a way that the advance pitch is substantially equal to or less than the projection of the energy zone between e2 and e3, that is to say by a distance corresponding to a6 in FIG. 3d. The range of energy levels between e2 and e3 makes it possible to transform into large-grain polychstal the material with small grains which was synthesized during the first irradiation (region a3 in FIG. 3b). In other words, this range of energy makes it possible to melt the layer of small grain material exactly up to the interface. As previously described, in this energy regime, there remain at the liquid / substrate interface some unmelted crystallites, from which solidification can take place. It will be noted that, since the latent heat of fusion of amorphous silicon is lower than that of crystalline silicon, the energy range e2-e3 is higher than the range e0-e1 which made it possible to obtain the same effect of coarse-grained crystallization in the starting amorphous silicon. It will also be noted that the zones e0-e1 and e2-e3 are energetically disjoint.

We will now explain in detail the different microstructures obtained in Figure 3d, depending on the energy of the beam and the pre-existing microstructure, that is to say created during the first irradiation. The coarse-grained region synthesized during the first irradiation (region a2 in FIG. 3b) is not completely remelted, because the energy density is never sufficient below e2 (it is recalled here that the threshold total melting energy of the polycrystalline material with reminiscence of scattered seeds which allows coarse grain growth is e2 and that the total fusion threshold which leads to homogeneous nucleation of small grains is e3); there is superficial fusion, but the microstructure of the polycrystal remains unchanged, because during resolidification, there is epitaxy of the liquid on the unmelted part of the underlying grain and practically the size of the grains remains invariant. The region denoted a5 in FIG. 3d corresponds to the irradiation below e2 of the small-grain polycrystal layer. There is partial fusion and regrowth from the original crystallites. During this regrowth, some grains will grow faster (depending on their crystallographic orientation) and a stratified microstructure is obtained, with larger grains on the surface. Zone a6 corresponds to the “SLG” regime for the small grain polycrystal (between e2 and e3) while above e3, there is still total re-melting of the small grain polycrystal (including the interfacial germs) and then, solidification by homogeneous nucleation, which again leads to a material with small grains. It will be noted that the region a4 in FIG. 3b has been completely remelted and transformed into small grain material.

FIG. 3f schematizes the situation after the energy beam has been translated by an additional step; for the reasons given above, the parts a5 and a6 of FIG. 3d are only partially melted and their microstructure does not change significantly. On the other hand, the material with small grains located beyond the zone a6 in figure 3d is now lit between e2 and e3 and thus crystallizes out of coarse-grained polycrystal. The area a6 in figure 3d has therefore been enlarged, to give the area a7 in figure 3f. In addition, the marking area of the front flank of the laser pulse of FIG. 3d has been redesigned and there is therefore no longer any marking problem.

Finally, Figure 3h schematizes the situation after a fourth pulse, still shifted by an additional step. It can be seen that the coarse-grained area a7 in FIG. 3f has been further enlarged, transforming into area a8. It is therefore possible to obtain, by using an energy ramp beam, a homogeneous material with large grains, and this, from the second pulse. In practice, to make electronic circuits, we will begin to use polycrystalline material on the glass plate only from the start of zone a6, zone a5 introducing grain size heterogeneity.

It will be noted here that the principle of the invention can also be applied to a starting material which is no longer amorphous, but polycrystalline or even microcrystalline; the principle, which in this case is simpler, is described in Figures 4a and 4b. As above, the energy range required to transform the poly or microcrystalline material into coarse-grained material is between e2 and e3. After the first pulse (FIG. 4a), the starting material is transformed into coarse-grained material in the region b2 corresponding to the energy range e2-e3. The energy is just sufficient to completely melt the thickness of the original material, while retaining some germs at the interface, from which crystallization takes place ("SLG" regime). In the regions b1 and b4 (flanks of the pulse), there is partial fusion and a stratified structure is obtained after solidification, with larger grains on the surface. Finally, in region b3, the fusion was complete, eliminating the residual crystallites at the interface, solidification takes place by homogeneous nucleation in the liquid phase and a homogeneous material with small grains is obtained.

In FIG. 4b, the beam has been translated with a pitch substantially equal to or less than b2. It can be seen that the small grain area illuminated in an energy range between e2 and e3 crystallizes from large grain material, while the defective b4 area has been completely remelted, giving way to homogeneous small grain material. The coarse-grained area has therefore grown and the “front flank” marking has disappeared. The use of an energy ramp also makes it possible, to a certain extent, to compensate for the energy drifts of the pulses delivered by the laser and to control the operation of the laser and / or of the beam scanning system (or else of the XY mobile table supporting the plates to be crystallized). This will be better understood on reading Figures 5a to 5f and their comments.

First, with reference to Figure 1, remember that the transition between “SLG” material and small grain material is very steep (the right part of the curve in Figure 1 falls very quickly), a phenomenon that provides contrast easily exploitable optics. In other words, the phase transition between the large material and the small grain material produces a marking on the plate during crystallization, which can be followed optically, using for example a video camera and a cathode-ray screen or the like.

There is shown in Figures 5a, 5b and 5c a normal pulse and its impact on the layer being crystallized in section view (Figure 5b) and in top view (Figure 5c). It is observed in practice that the border FR1 between the zone of coarse-grained material ZO and the zone of small-grained material Z1 is very clear and extends over a few microns barely, on a microscopic scale. In FIG. 5d, after a displacement of one step, the normal pulse In1 has been shown in dotted lines, as well as the impact which it should have provided on the layer (normal SLGn zone); a pulse with a lower average energy level has also been represented (in solid lines) representing a real pulse Ir which in reality provides an offset SLGr zone and therefore leaves in the layer an zone Z2 of small grain material located between the zone ZO in coarse-grained material and a new zone Z3 in coarse-grained material. Suppose now that a MAR mark has previously been registered on the real-time observation screen, a mark corresponding to the alignment of the border which should have been obtained between the coarse-grained material and the small-grained material . The marking FR2 produced by the pulse at lower average energy will no longer be aligned with the fixed mark on the observation screen and the offset D1 (see FIG. 5f) is perfectly measurable on the screen. This observation can be made by an image processing system. It is therefore sufficient, knowing D1, to command the retraction of the laser spot by an amount D1 on the next pulse to correct the "hole" of crystallization in small grain material which had been left during scanning. We thus introduced a feedback loop providing a feedback signal which allows the beam scanning to be controlled in real time. This enslavement can also lead to modifying the displacement step of the movable table. It is noted that one could also control the discharge voltage of the laser in order to correct the average energy per pulse.

FIG. 7 represents a system making it possible to implement the method of the invention. This system comprises between the source S1 and the layer of semiconductor material an absorption device Ab. device is, for example, an absorbent strip having a bevel profile so as to provide an absorption gradient along the axis OX.

For example, a slope at 20%, for an “SLG” zone at ± 5% of the width of the laser beam, for a recovery rate of 98 to 99%, gives a uniform polycrystalline silicon material of a size average grain size from 0.8 to 1.1 micron in diameter.

Even weak draw-to-draw stability (eg ± 1%) in a top-hat beam profile configuration will generate random markings of the irradiated material. The ramp eliminates this drawback. In fact, it suffices to adapt the slope of the ramp to the laser-to-laser stability. Knowing that 10% of the surface of the ramp width is crystallized of the “SLG” type, it suffices to choose the slope of the ramp as a function of these instabilities. Strong instabilities will therefore preferably involve the use of a strong ramp.

Any improvement in the laser-to-laser stability allows the use of a less steep slope (figure 8) as a result, the “SLG” zone is wider and the use of a lower recovery rate is possible, ie saving crystallization time at constant laser frequency.

According to variant embodiments shown in FIGS. 6b and 6c, provision can be made for the pulse to present, upstream of the pulse (in the direction of movement of the pulse) a decreasing energy level profile comprised between e4 and e5 in zone a7 which allows an in situ dehydrogenation of the material just before its crystallization by the previously defined pulse. This eliminates an external technological stage (thermal annealing for several hours in an oven). The invention as described above thus makes it possible to achieve, on a surface made of amorphous semiconductor material, the crystallization of a strip of this material by displacement of the impact of a laser spot on this surface. As shown in FIG. 9, a first movement of the laser spot along arrow D1 makes it possible to produce a first band B1 of width I. To carry out the crystallization of a whole surface whose width L is greater than the width of the band B1, it is necessary to crystallize several bands side by side. After having made strip B1, a second strip B2 is therefore made, etc. However, the b12 border between two bands may be of poor crystallographic quality. According to the invention, the laser impacts produced for the manufacture of the strip B2 overlapping the strip B1. The energy ramp profile as described above makes it possible to remelt the crystallized material of the band B1 and to recrystallize it.

Referring to Figure 10, we will describe this crystallization process.

In this figure, the band B1 is shown produced by a first displacement of the laser impact. Each impact such as the Xth impact has two lateral flanks f1 and f2. The impact has, in the direction parallel to the width of the strip, an energy profile as shown in the upper part of the figure and, in the direction of movement B1 of the impact, a ramp profile as shown in Figures 5a-5f, or irregular (Figures 6a-Tc).

In the band B1, the area "a" corresponds to an area of the sample having been irradiated by an energy gradient due to the flank f1 of the laser beam. This material will be composed of a mixture of small and large grain polysilicon. Since this corresponds to an unusable edge of the sample anyway, this does not matter. Zone b will be composed of a uniform coarse grain material by using the ramp profile. Zone c will be of the same nature as zone a, for the same reasons (right flank). Zone d corresponds to an amorphous silicon which has not yet crystallized.

During the second scan carried out parallel to the first scan and of which the ^same impact has been shown, the second strip B2 is produced. Impacts such as the ^same impact overlap the band B1 so that the flank f1 of each impact is entirely in the area b of the band B1. Zone c of the first scan as well as a small part of zone b will be completely recast thanks to the exposure to the energy level contained in the energy profile.

Zone b, large grains, will then be extended by the length of zone c 'to become a zone b'. The zone d will have been reduced to a zone of. A new zone c is created at the right end of this second scan. The zone of coarse-grained material b will have been enlarged by pushing the degraded zone laterally (zone c). This will be the case, with each subsequent scan, if necessary, depending on the width of the sample to be crystallized and the length of the laser beam.

By this method, we will then have uniformly crystallized a sample whose dimensions may be much greater than the width of the laser beam while eliminating the problems of beam splicing.

Application example

For a laser beam of 30 cm * 0.5 mm and an industrial laser operating at 300 Hz with a recovery rate of 98% (50 shots per unit of area) the step of 10 μm allows to crystallize 3 mm of material per second, or 100 s required to process a 30 ^* 30 cm screen. If a recovery rate of 95% is acceptable, then the processing of the same screen would only take 40 s. These times are industrially acceptable. We can act on the energy slope of the laser beam.

The use of a steep slope will tolerate large variations in energy from source to source in contrast to a softer profile which will require better energy stability since it draws from the laser.

This choice will depend

- energy accuracy from laser to laser

- the width of the coarse grain zone (“SLG”) produced

- the catch-by-sink recovery rate (number of strokes seen per unit of area)

In the foregoing, a light source is used as the energy source, but any other energy source, for example an electron beam, could be used.

Claims

1. Method for crystallizing a layer of semiconductor material (1) using an energy beam (F1) illuminating on a surface (L x I) the layer of semiconductor material (1), the crystallization of the layer by relative displacement of the beam on the surface of the layer in a determined direction (OX), characterized in that said beam has an energy profile in the direction of displacement (OX) such that there is at least a first energy level allowing to obtain a large grain crystallization (Super Lateral Grow - SLG) from an amorphous or small grain crystalline phase of the semiconductor material as well as one or more second energy levels higher than the first level of energy and located towards the front of the area illuminated by the beam in the direction of movement (OX) of the beam and allowing a fusion / dissolution of a material comprising both large grain crystalline phases e t small grain crystal phases.

2. Method according to claim 1, characterized in that at least a second energy level is geographically close to the first energy level and in that the energy beam is impulse so that after a first illumination of a first zone, the beam is displaced on the surface of the layer of semiconductor material and that a second illumination of a second zone induces sufficient energy downstream of the first zone to carry out the fusion / dissolution of the semiconductor material.

3. Method according to claim 2, characterized in that an energy pulse comprises a front flank having an energy gradient and in that the displacement of the beam between the first illumination and the second illumination is such that, during the second illumination, the beam portion having a second energy level illuminates the zone (a4) previously illuminated by the front edge of the pulse during the first pulse.

4. Method according to claim 1, characterized in that the first energy level allows the fusion / dissolution of a crystalline phase with small grains of the semiconductor material.

5. Method according to claim 2, characterized in that the energy beam is a light beam.

6. Method according to claim 5, characterized in that the energy beam has an energy gradient containing said first and second energy levels.

7. Method according to claim 6, characterized in that the energy gradient is linear.

8. Method according to claim 6, characterized in that the energy profile of the beam has an energy level corresponding to the first energy level preceded in the direction of movement of the beam by an energy peak corresponding to the second level of energy.

9. Method according to claim 6, characterized in that the front flank of the energy beam in its direction of movement has a decreasing energy gradient towards the front flank so as to cause heating of said layer for a time sufficient to allow its dehydrogenation.

10. Method according to claim 1, characterized in that it provides for viewing the crystalline zones with large grains and the crystalline zones with small grains and for detecting their borders (FR1, FR2); in that we determine according to the shape of the energy beam, the theoretical position (MAR) of each border and in that we compare the position of each border (FR2) with its theoretical position (MAR) so as to achieve a feedback system deriving from the distance separating the theoretical position (MAR) from the position of said boundary (FR2), a feedback signal acting either on the position of the laser beam, or on the displacement in said direction (OX ) of the layer of semiconductor material (1) to be treated, or on the discharge voltage of the laser to correct any discontinuity in the coarse-grained material due to energetic drifts of the pulses delivered by the laser.

11. Method according to one of claims 1 to 10, characterized in that it comprises a first displacement (D1) in said direction (OX) so as to achieve crystallization of a first band, then at least a second displacement (D2) parallel to said direction, the illumination of the layer of semiconductor material during this second movement overlapping the first strip which was illuminated during the first movement so that a lateral flank located on one side of the beam of illumination perpendicular to the direction of movement of the beam illuminates the crystalline zone with large grains of the first displacement (B1).

12. A crystallization system for a layer of semiconductor material comprising a light source illuminating, using a light beam, one face of the layer of semiconductor material as well as means for moving in a determined direction and direction the light beam with respect to the layer of semiconductor material characterized in that it comprises means (Ab) between the source and the layer of semiconductor material for inducing an energy gradient of the level.

13. System according to claim 12, characterized in that said means located between the light source and the layer of semiconductor material comprise an element operating in transmission and having an absorption gradient.