CA1129969A - Dual wavelength optical annealing of materials - Google Patents

Abstract

15.
DUAL WAVELENGTH OPTICAL
ANNEALING OF MATERIALS

Abstract A new mode of radiant heating is disclosed wherein two different wavelength pulses (FIG.
7, 42, 43) are used to melt portions of a solid body (41). The first pulse may be of relatively weak intensity, but is strongly absorbed by the solid. The second pulse, which is not strongly absorbed by the body when in the solid phase, is strongly absorbed by the body when in the molten phase. Exposure to the first pulse results in an initial melting of the body, which then becomes highly absorptive to radiation at the wavelength of the second pulse. Readily available radiant energy sources, e.g., lasers, which generate radiations generally not highly absorbed by the body in the solid phase may thus be efficiently utilized.

Description

` ~?~3~

1~
DUAL WAVELENGTfl OPTICAL
AN~læALING OF MATERIALS

Backg_ound of the Invention 1. Field of the Invention _ __ This invention involves optical heating of material, and particularly to optical annealing of semiconductor material.

2. Description of the Prior Art 'rhe electrical properties of semiconductor materials are engineered to Eulfill specific electrical require~ents by doping the pure starting material with appropriate constituents which alter the conduction properties of the semiconductor. Such doping constituents 15 are added to the starting semiconductor primarily by either the diffusion or the ion implantation process. These processes, especially the ion implantation process, may result in significant damage to the crystal structure of the semiconductor. Such damage renders the resultant 20 material ineffective as an active element for most electrical applications~
In present day fabrication processes, the damage incurred during the doping step is repaired or "annealed"
by bringing the material to an elevated temperature. The 25 increased mobility and diffusion characteristics at these elevated temperatures allo~s both the host and the dopant constituents to reorient themselves so that a more perfect crystal structure results with concomittant electrical properties that render the devices electrically operative~
30 Annealing is also used in other situations, described hereina~ter, not involving damaged materialO Prior art annealing techni~ues have involved the simple use of an appropriate oven to raise the damaged material's temperature as required.
Recently, the laser has been effectively applied to this annealing process. E~posure of the damaged semiconductor to laser radiation results in increased mobility and diffusion rates necessary for effective . ~ ~

:
:

' . .

annealing. How~ver, unlike the prioe art ther~al anneal ng process, two specific and identifiably distinct regimes are found to occur in the laser annealing process. In the first regi~e, the temperature of tlle substrate is elevated 5 in a manner similar to the prior art thermal process. As in the prior ar~ process, the semiconductor retains its solid phase tllroughout this "solid phase epitaxial regrowth" regi~e.
A second laser annealing regime, without 10 comparable precedent in the prior art annealing process, involves operating under parameters which result in the melting of that part of the substrate which is exposed to the laser. The di~fusaon rates and mobility in the molten phase are significantly different from that in the solid 15 phase and, as a result, this annealing regime has radically difEerent physical and temporal characteristics than ~he prior art process. When the substrate is no longer exposed to the laser energy, the molten region refreezes to a crystal, using that part of the underlying undamaged 20 crystalline semiconductor material which is not melted as a template or seed from which to regrow. The process is consequently referred to as "liquid phase epitaxial regrowth" annealing.
The high powers which are necessary for laser 25 annealing are currently most readily available in devices which emit radiation in the infrared region of the spectrum. However, most semiconductors are only weakly absorptive in this region oE the spectrum, and hence, laser annealing with infrared sources is found to be inefficient, 30 if at all possible, with such sources. On the other hand, while most serniconductors are found to absorb visiblé light effectively, currently available laser sources in this region of the spectrum have only lilnited power and are consequently inefficient for commercial laser annealing 35 applications~
: `
''`

' ,~,, ,,~

, : .

.

2a.

Summary of the Invention In accordance with an aspect of the invention there is provided a process for heat.ing a solid body to render a portion thereof molten by exposing the body to electromagnetic radiation, characterized in that the radiation is of two different wavelengths, the first wavelength radiation being absorbed by the body and being effective to initiate melting of said portion, and the second wavelength radiation being more efficiently absorbed by the body in its molten state than in its solid state and being effective to extend the molten zone.
This invention is an improved technique for radiant heating of materials, such as semiconductors.

: ' .~

. . .
-:

,, .
n~

3.
The technique involves exposing the material to a first pulse, usually of relatively lo~ intensity radiation, which is strongly absorbed by the material. This initial exposure results in the melting of at least a thin 5 surface layer of the material. This molten layer is then exposed to radiation which would not normally be strongly absorbed by the material when in the solid phase, but which is strongly absorbed by the material when in this molten phase. As a result of this invention, the 10 practitioner may make effective use of those radiant energy sources which emit radiation that would be relatively ineffective in heating the material when in the solid phase. At the same time, only minimal energy is required from those radiant energy sources which are 15 highly absorptive by the solid material. The invention makes optimal use of these distinct sources in those ~` specific operative regimes where they are most highly and uniquely effective. In a specific embodiment, a single wavelength laser is used in conjunction with a frequency 20 altering device, such as a frequency doubler, to yield two different wavelength portions.
Brief Description of the Drawing i Figure 1 is a schematic representation of an apparatus useful in the practice of this invention.
Figures 2-4 schematically~ represent reflectivity data which demonstrate the particular efficacy of the dual wavelength laser annealing process.
Figures 5-~ are plots of the melting ~ properties of various semiconductors as a function of ; 30 energy of the impinging light for various wavelengths.
Figures 7 and 8 are schema~ic representations of specific embodiments of the invention.
Detailed Description The invention is an improved technique 35 for radiantly heating materials, e.g., for the purpose of annealing semiconductor materials. The invention stems partly from the realization that the radiation absorption characteristics of materials depend significantly upon - ,~

,. ~ .. .. .

:j , .
, ......... . .
,' ' ' . .

,.3~?~

whether the material is in the molten or solid phase.
For example, readily available and high-powered infrared sources are not generally effectively absorbed by the exemplary semiconductor materials in the solid phase. On 5 the other hand, low-powered visible wavelength sources are effectively absorbed by the solid phase semiconductor. The invention involves utilizing a pulse of radiant energy which is strongly absorbed by the materials to render at least the surface of the material 10 molten. Another pulse of radiant energy, in a region of the spectrum which is not strongly absorbed by the ;~ material in the solid phase, but which is strongly absorbed by the molten phase, is then used to yield effective annealing.
The required thickness of the molten region which is created by exposure to the first pulse is dictated by the absorptive phenomenon which takes place during exposure to the second pulse. The physical characteristics of the molten phase absorption are such that the molten region 20 created by the first pulse will usually have a thickness approximately equal to the optical skin depth of the material at the wavelength of the second pulse. In many applications, this molten regime will display properties commonly referred to as "metallic!'. However, this 25 invention is concerned primarily with the optical absorpti~e properties of the molten regime and is less concerned with some of the other properties associated with metallic materials.
; Figure 1 is a schematic representation of an 30 apparatus which may be used in the practice of this ~;~ invention. In this figure, a beam of light 11 i5 emitted by a source 10, onto a substrate 12. The light source 10 produces the two wavelengths of light necessary ~ for the practice of this invention and directs - 35 them with appropriate timing characteristics to the substrate 12~ he source 10 may simply be two standard light sources of sufficient power and possibly with - appropriate filters~ or two lasers of appropriate ~,,, .,.. ~ ",~".1 . :. :
:~:. .: . : .
~. ~ , .. .. .

~ ?~

s .
~avelengths. In a specific embodiment, the source is a single laser with an appropriate frequency doubling crystal.
The various parameters involved in laser 5 annealing may be effectively monitored by utilizing a known reflectivity technique. In this technique, the molten phase is monitored by observing the light reflectivity of the substrate surface in response to the application of a pulse of laser energy thereto. The 10 technique depends on the increased reflectivity of the semiconductor when in the molten phase as compared to its reflectivity when in the solid state. The data in Eigures 2-4 was obtained using that technique. The figure represents the results of an experiment involving 15 laser annealing of a silicon substrate implanted ~; with arsenic atoms, the implanted region of the substrate thus having been damaged by the implantation process.
Figure 2 displays the reflectivity of a substrate exposed to a 3.5 joules per cm2 pulse of 1060 nanometer 20 radiation. In the figure, the vertical axis indicates the reflectivity of the semiconductor surface exposed to the laser light. The horizontal axis represents time. It ~ is apparent from figure 2 that, for silicon, the 1060 nm -~ light is not strongly absorbed. The lack of reflected ~; 25 light from the substrate, as indicated by the essentially horizontal data plot, shows that no melting of the semiconductor took place. In figure 3, an identically - prepared substrate was exposed to a pulse of 530 nm light ` at an energy density of 300 mj per cm2. In this figure, ~` 30 it is apparent that the substrate surface has been melted, as shown by the enhanced reflectivity of the substrate surface. While the absorption characteristics l of the substrate increases radically from 10~0 to 530 nm, ¦ yielding the observed molten phase, in this case the 35 energy available from the source at 530 nm was not - adequate to melt the substrate to sufficient depths(as - indicated by the length of time the substrate surface is , molten)so as to reach the underlying single crystal . . ~

. ~ ~, . . .
.~ -, , .. .

material. Consequently, the processed m3terial is not single crystal and is not adequately annealed. Figure 4 illustrates the results when an identical substrate is exposed to a pulse of 1060 nm light and 530 nm light with 5 energies as in Figures 2 and 3. In this example, the semiconductor is not only rendered molten, but remains molten for a much longer period than that obtained in figure 3. This synergistic effect results since the - material rendered molten by the 530 nm light becomes 10 hiyhly absorptive to the high-powered 1060 nm light to which it was previously essentially transparent.
Essentially, the full power of the 1060 nm light is then absorbed by the molten region extending the depth of this region further so that part of the single crystal 15 undamaged material underlying the damaged layer is also rendered molten. As a result, when the laser is shut ~` off, and the molten material recrystallizes, the single crystal underlying region acts as a seed or template from which the molten regime may regrow in single crystal ~:! 20 form~
~` It may be seen from the above discussion that the absorptive characteristics of the semiconductor at various wavelengths is a critical aspect of this I invention~ ~he invention may be effectively practiced as 25 long as the solid semiconductor is sufficiently ` absorptive at the wavelength of the first light pulse so `l that a melt results. This pulse will consequently be referred to as the "melt initiating pulse".
It should be noted that there may be 30 various regions of the electromagnetic radiation spectrum for which the solid semiconductor is absorbing. In general, visible light is absorbed strongly by most semiconductors as a result of prevalent electronic transitions. However, strongly coupled absorption by the 35 solid semiconductor may also occur in the far infrared region of the spectrum as a result, for example7 of vibrational excita~ion. In such a case, the wavelength of the "melt initiating pulse" may be in the far infrared j,~
i ,~:
: , , : , ., ~; , .
:, ~

region of the spectrum. With other materials which may be processed according to thisinvention, still other wavelength radiations may be found useful.
Once the surface layer of the 5 semiconductor has been rendered molten, the second pulse can couple with this molten region which is highly absorbent to its light. Light from this second pulse, since it is now readily absorbed, will generally extend the melt to greater depths. This pulse is consequently 10 referred to as the "drive in" pulse.
The physical characteristics of both the melt initiating and drive in pulses are determined, in this embodiment of the invention, by the requirement that exposed portions of the semiconductor be rendered molten 15 to a depth sufficient to include some of the underlying single crystal material. This requirement governs the ; temporal and energy characteristics of the pulses, as well as their wavelength. These characteristics are displayed in part in Eigures 5 and 6 where the energy of 20 a 40 nanosecond pulse is plotted on the horizontal axis, and the duration of the resultant molten phase is plotted on the vertical axis. Fig. 6 uses an expanded scale, in comparison with Fig. S, for more accurately illustrating the relationships at low pulse energiesO Curve 31 is for 25 a gallium arsenide substrate and laser light of 530 nm.
Curve 32 is for a germanium substrate and a wavelength of 530 nm. Curve 33 is for a silicon substrate exposed to 530 nm light. Curve 34 is ~or a silicon substrate implanted with 30 kV arsenic atoms to a density o~ 1015 30 atoms per square centimeter and exposed to 530 nm light.
Curve 35 (Fig. 5) represents a silicon substrate exposed -- to 1060 nm light. Curve 36 represents -~ silicon implanted with 30k~ arsenic atoms to 1015 atoms per square centimeter and exposed to 1060 nm light. It is 35 clear from Fig. 5 that the shorter wavelength light is much more readily absorbed by the crystal than the longer wavelength light. Consequently, foe this shorter wavelength light, longer melt durations are obtained with ~ `"'' , ~, ' ; , ' ' ' ` ,,:
: ;~ ,, ,, , ;, ,: ' - ' 8.

lower energy deposition. While molten conditions may be obtained using the long wavelength light, the figure shows the inefficiency of such a process ~hen compared to the use of the shorter wavelength light to initiate the molten 5 phase. Additionally, it should be noted that the ; efficiency of the longer wavelength light in rendering the crystal molten is highly dependent upon implant dose. In contradistinction to this, the efficiency of the shorter wavelength light in rendering the crystal molten is much 10 less dependent on implant dosage. Once the crystal is rendered molten, it becomes highly absorptive over a broad range of the spectrum with an absorptivity which is essentially independent of implant dosage.
The temporal extent o~ the laser pulses is ~ 15 determined by the requirement that the substrater be ;l rendered molten. As such, both the absorptivity of the material and its thermal diffusion must be considered in determining the optimum length of the pulse. The absorption o~ light by the material must be sufficiently ;~ 20 rapid so that the heat will not be lost as a result oE the ~ thermal diffusion. Hence, the absorptivity must be high `~ enough relative to the thermal diffusivity so that melting will occur. In the experiments performed by applicants, 5-100 r~s pulses were found to be most efficient. However, - 25 the pulse width will be materially dependent on ~he nature -~ of the material, its physical characteristics and the wavelength of the light. In many applications, the pulse width will be less than 150 ns and in most applications, the pulse width will be less than 1 microsecond. These 30 considerations are equally important for both the melt initiating pulse and the drive in pulse.
}~ The time delay between the first and second pulses may be critical. For example, if the second drive in pulse arrives after the first 35 melt-initiating pulse has been extinguished, the molten region may have refrozen and the drive in pulse may be ~;f totally ineffective. Likewise, if the drive in pulse arrives prior to the melt-initiating pulse, that portion , . .
, :

.:
`: ~: ' , ,. :
: ~, ~' "

of it which appears beore the arrival of the melt-initiating pulse will be ineffective~ The most efficient operation occurs when the melt-initiating pulse arrives just before the drive in pulse and overlaps with 5 it. The nomenclature used herein referring to the pulses as the first pulse and the second pulse, does not, however, refer to the temporal relationship between the pulses, but refers rather to the functional effect of the pulses. The first melt-initiating pulse renders the 10 surface molten and the drive in pulse strongly couples with this molten region. However, as discussed below with reference to Figure 7, in some embodiments, the "second" pulse may arrive before the "first" pulse.
The depth to which effective annealing 15 can proceed is in part determined by thermal diffusion considerations in the melt. The thicker the molten region, the greater distance over which the heat must diffuse to extend the molten region. For depths greater than one micron, the surface, which is the hottest, will 20 begin to display physical damage. The embodiment of Figure 8, discussed more fully below, is then ~` particularly advantageous, since in this embodiment the ` drive in pulse irradiates the substrate from the opposite side of the melt-initiating pulse. In this 25 configuration, the heat is absorbed at the melt-solid interface and need not diffuse by means of a thermal gradient. Hence, greater depths may possibly be rendered ; molten without surface damage.
~- The pulses required for the practice of 30 this invention may originate, for example, from two separate laser sources. However, a most effective way oE
~-- practicing this invention is through the utilization of a single laser which emits light at the wavelength of one of the pulses, and then transformation of a part of this 35 pulse, in a non-linear device, into light at the wavelength required for the other pulse. In a specific embodiment, an ;~ Nd-glass laser emitting light at 1060 nm may be partly frequency-doubled to yield light at 530 nm. In this ~, i ~.

, ~ .
, : , ~`,- : ' ;, :

, 10 .
specific embodiment, the 530 nm pulse comprises the first rnelt initiating pulse, while the 10~0 pulse comprises the second drive in pulse. In other embodiments, the laser source may emit light which comprises the first pulse, and the non-linear processing may yield light which comprises the second pulse. The light pulses may be polarized or unpolarized and may be co-linear or may impinge on the substrate from different angles of incidence.
Substrates which can be processed according to 10 the teachings of this invention can be of doped semi-conductor materials. As discussed abover the doping process may involve any one of the prevalent semiconductor doping processes. A portion of the semiconductor material which is damaged is then exposed to the annealing process 15 so as to anneal the damage and render the semiconductor electrically active. ~nderlying the damaged region is a substantially perfect single crystal material to which the molten region must extend in order for effective annealing to occur. Usually, the single crystal material will be 20 identical to the original starting semiconductor material - before it was damaged. For example, the starting material may be a single crystal silicon wafer which is damaged to some depth during ion implantation. The underlying base silicon remains single crystal and it is this material 25 which provides the seed for regrowth during the annealing processO However, in some embodiments, the underlying ~l single crystal material may not be exactly identical to the damaged material, but may be sufficiently analogous to its desired structure so that it may effectively act as a seed for regrowth. For exampler a sapphire substrate will act as an effective seed so that overlaying silicon may be laser annealed according to the teachings of this invention.
While many embodiments will involve the ~^ 35 annealing of implanted semiconductor material, in some embodiments, annealing is performed for other purposes.
v~ For example, silicon may be deposited on a thin single crystal substrate using, for example, an evaporative :-$

. "

. , '. . .;. . . ~ ;. . .

, . , ` `, ,. . , ~.

' ` ` ` . . ` . .. .

, . ~.: ` . . .

., ., ~ . ~` .
.

process. The deposited silicon is then amorphous, but may be rendered crystalline by the annealing process described above. The instant invention is useful to perform such processing, and may further be used to 5 anneal other materials, such as a surface layer of dopant, which are either amorphous or lack perfect crystal structure. Exemplary semiconductor materials include silicon, germanium and gallium arsenide.
While the effects of the annealing 10 process on impurity distribution have not been discussed, they may be significant, and may be advantageously utilized. In some situations, the impurity may be implanted near the surface and upon melting, may diffuse throughout a significant portion of the depth of the 15 molten region. As a result, the drive in pulse which is used to increase the depth of the molten region may also be used to increase the depth to which the implanted impurities diffuse.
The nature of the second pulse, ~;20 specifically the fact that it is not strongly absorbed by the solid phase material, allows one to practice this invention in a number of specific configurations which ;permit more convenient laser annealing than heretofore possible. For example, figure 7 shows a piece of 25 semiconductor material 41 portions of which the practitioner desires to anneal. The entire substrate is exposed to drive in light 42. Because of the nature of this light, and the fact that it is not absorbed by the solid phase, the light has little affect on ~he substrate.
30 However, portions 44 of the substrate are additionally exposed to melt~initiating light 43~ Those regions of the substrate which are exposed to the melt-initiating light are rendered molten, and are thereby made capable of absorbing the dri~e in light, 42. Upon 35 absorption of this light, the regions 44 are effectively annealed. This particular configuration allows one to use ~;the melt-initiating pulses to define specific regions which are annealed, while the entire substrate is exposed to the , ,, ~
, ~ ' ' ,,~, ~

12.
drive in light.
In another embodiment, shown in figure 8, a substrate, S0, with a damaged region, 51, is exposed to drive in light, 53, from its far side. It should be noted 5 that in this embodiment, the practitioner is taking advantage of the fact that this substrate in its solid phase is transparent to the drive in light. The damaged region, 51, is exposed on its near side to melt-initiating light, 52. Those regions of the substrate which are 10 exposed to the melt~initiating light, and identified as 54, will then be rendered absorptive to the molten phase light and will be annealed.
As discussed above, this embodiment is particularly advantageous since the thermal energy is 15 deposited by the drive in pulse at the melt solid interface and need not be diffused from the surface to this interface by means of a thermal gradient. Consequently, physical damage which is observed in the previous embodiments when the molten region becomes too thick is not observed in this Z0 embodiment.
EXAMPLE
An arsenic doped silicon wafer was annealed using the above-described process. The silicon wafer was three inches in diameter and was ion implanted with 30 kV arsenic 25 atoms to a depth of 300 Angstroms and a density of 1015 arsenic atoms per square centime~er. Subsequent to implantation, the 500 Angstrom damaged layer was annealed by exposing the surface to a frequency-doubled Nd-~lass laser. The laser apparatus included a Q-switched 30 Nd-glass laser, a Nd-glass single pass amplifier, and a KDP crystal frequency doublerO The system emitted 3.5 joules per cm2 1050 nm light and .3 joules per cm2 frequency doubled 530 nm light. The infrared light beam was 8 millimeters in diameter and the associated pulse 35 had a 40 ns full width at half maximum, while the visible ; light had a 6 millimeter diameter with a 30 ns full width at half maximum. The pulses were co-linear and were impinged on the wafer at a 10 degree angle to prevent .. , : . . . . . .
' ~ ' `,: ' ' ., . '., ' ' , ' , '., ' ':' ' .
.

13.
reflection of the light into the laser. Reflection studies from the silicon surface indicate that the molten phase existed ~or approxi~ately 2~0 ns and extended into the substrate approximately 2500 angstroms. The damaged 5 region which was exposed to light was effectively annealed as determined by subsequent Rutherford back scattering and channeling studies.

~':
:

.
~`.

~'~

~

, ................ . .

. ~ ~
~ .. ,,.. ~ .

'.,~ ~ ~ ` . . ' . , , ~ .~ ' .
' l .

Claims (6)

Claims:

1. A process for heating a solid body to render a portion thereof molten by exposing the body to electro-magnetic radiation, CHARACTERIZED IN THAT
the radiation is of two different wavelengths, the first wavelength radiation being absorbed by the body and being effective to initiate melting of said portion, and the second wavelength radiation being more efficiently absorbed by the body in its molten state than in its solid state and being effective to extend the molten zone.

2. A process according to claim 1, CHARACTERIZED IN THAT
the radiations are provided in pulses.

3. A process according to claim 2, CHARACTERIZED IN THAT
said first and second wavelength radiations are provided as separate pulses overlapped in time.

4. A process according to claim 1, CHARACTERIZED IN THAT
said two wavelength radiations expose different but overlapping portions of said body, only said overlapped portions being rendered molten to said preselected depth.

5. A process according to claim 1, CHARACTERIZED IN THAT
a single laser source is used, part of the output radiation thereof, at said first wavelength, being transformed to said second wavelength.

6. A process according to claim 1, CHARACTERIZED IN THAT
said portion is exposed directly to the first wavelength radiation, and is exposed to the second wavelength radiation through said body.