semiconductor layer transfer by hydrogen implantation in combination with wafer bonding technology has opened new possibilities for the fabrication of silicon-on-insulator materials such as silicon-on -oxidized silicon, silicon-on-quartz, or similar structures such as SiC-on-glass or GaAs-on-silicon.

For the purpose of layer transfer, splitting of whole wafers parallel to the bonding interface along hydrogen-induced microcracks is required.

However, for the understanding of the processes leading to these microcracks, it is much more convenient to investigate the development of surface blisters on hydrogen-implanted but unbonded wafers.

Tong et al. studied the surface blistering process as a function of annealing temperature and annealing time in various materials by optical microscopy.

As a characteristic quantity to describe the blistering process,the annealing time required for the formation of optically detectable surface blisters was used.

This characteristic time which showed an Arrhenius relationship with annealing temperature was termed "onset" or blistering time.

Of course, the question then arises whether this onset time does have a specific physical background.

The question may be formulated as follows: Do hydrogen-induced microcracks pop up as surface blisters at the onset time after originally closed hydrogen-induced platelets have grown to critical size, or is this experimentally defined onset time for blistering just a parameter associated with the resolution of optical microscopy?

In the present letter we report on investigations which answer this question based on optical microscopy combined with atomic force microscopy (AFM)

4 in., (100), Czochralski,p-type boron doped silicon wafers covered with 150nm SiO2 were implanted by H2 with a dose of 5*10/cm2 for different implantation energies of 38, 90, and 129keV.

After implantation, the SiO2 layer on some of the wafers was removed by wet etching to get different thicknesses of the later over the hydrogen implantation peak.

This way, we got a group of hydrogen-implanted silicon wafers with different hydrogen implantation depths as shown in Table1

The H2+ implanted silicon wafers were then cut into small pieces in the size of 2mm*2mm and annealed in air at different temperatures.

The onset time for surface blistering at every annealing temperature was determined as the annealing time at which tiny blisters can first been observed on the sample surface at a magnification of 500 under an optical microscope with a differential-interference-contrast system.

AFM was used to overcome the resolution limit of optical microscopy and, therefore, to check whether surface blistering already occurs before these blisters could be detected optically.

The size of surface blisters, when present, was also determined by AFM.

By optical microscopy, we found that surface blisters do not form if the annealing time has not reached a critical value.

However, upon reaching the onset or blistering time, all of a sudden, tiny blisters in the size of 1-2um depending on hydrogen implantation depth appear on the sample surface.

With further annealing, these blisters grow and finally break.

In the case of bonded wafers, the blisters may merge and cause splitting over the whole area.

In order to ensure that the observed sudden appearance of small blisters is not just caused by the resolution limit of optical microscopy, we used AFM to study the surface morphology of five silicon samples which were implanted by H2 and annealed afterwards at 350℃ for 5, 8, 12, 15, and 17s, respectively.

17s is the onset blistering time at 350℃ determined by optical microscopy.

The size of blisters, when present, is measured more precisely by AFM and is considered zero if surface blisters are not detected even by AFM.

The corresponding blister size versus annealing time is given in Fig1.

The surface images obtained by AFM for the 15s (before the onset blistering time) annealed sample and the 17s (just after the onset time) annealed sample are shown in Fig2.

The AFM results above clearly show that the occurrence of these tiny blisters is more like a sudden pop-up step at a critical time and not a gradual growth process.

The AFM results thus rule out the possibility that the onset time is just caused by the optical resolution limit and has no actual physical significance.

To summarize our experimental investigations, the onset of blistering during thermal annealing in hydrogen implanted materials is schematically illustrated in Fig3.

For the purpose of simplicity, it does not take into account the growth of larger platelets at the expense of smaller platelets, i.e., Ostwald ripening.

The onset of blistering was also investigated from a theoretical point of view.

Hydrogen implantation initially leads to the formation of hydrogen containing platelets on (100) or (111) planes.

During annealing, the diffusing hydrogen may enter the platelets and form hydrogen molecules associated with a certain pressure which forces the platelets to grow.

In the very beginning, the growth is more likely in a lateral propagation manner since the platelets themselves are not yet open, but hold together via van der Waals forces between the two hydrogen-covered internal surfaces of the platelets.

The parallel extension of the pressured platelets in hydrogen-implanted silicon as well as in other materials, e.g., Ge, is evidenced by high-resolution transmission electron microscopy TEM.

Based on the TEM results in literature of the hydrogen-induced platelets/microcracks and our AFM results which show that up to the onset time the buried hydrogen agglomerates do not bulge up, we conclude that the platelets grow in a closed form up to a critical size at which the internal pressure is high enough to overcome the surface energy γp associated with the interaction between the hydrogen-covered surfaces of the platelets.

It can reasonably be assumed that γp is much lower than the surface energy γsi of the corresponding crystal planes of silicon determined by cleavage method.

An expression of a critical radius derived by Mitani and Gosele, for an analogous situation (the nucleation of interface bubbles formed between two bonded thin wafers) may also be used for the onset of blistering in hydrogen implanted silicon.

In Eq(1), Δp is the difference between the pressure inside the platelets and that of the outside atmosphere, t corresponds to the hydrogen implantation depth, E is Young's modulus, ν Poisson's ratio, and α a numerical factor in the order of -1 depending ont the details of the calculation.

It is worth pointing out here that we only use Young's modulus for silicon in our case though the top layers of some samples contain both Si and SiO2 layers of comparable thicknesses (see Table1).

Since Young's modulus of Si is about twice that of SiO2 , the simplification we used in our treatment may cause some discrepancy between experimental results and theoretical predictions which will be presented later.

Therefore, we use Eq(1) to give a qualitative rather than a fully quantitative description of the onset of blistering.

In agreement with our experimental observations that the onset blisters always come with a certain size, Eq(1) confirms the existence of a critical size of the growing platelets before they pop up as surface blisters.

If we assume γp=0.5J/m, E=1.3*10Pa, t=500nm, α=1, ν=0.3 for silicon, Δp=10-10Pa, we get r inthe range of 0.4-1um which is in good agreement with the onset blister size (-1um) determined experimentally.

Equation(1) also predicts that an increased implantation depth or top layer thickness will lead to an increased critical size of the fully grown platelets (also the size of the first emerged surface blisters).

The calculated results based on Eq(1) and the AFM measurement of the layer thickness dependence of the onset blistering size are given in Fig4.

The theoretical curve qualitatively agrees with our experimental results though the experimentally observed increase is smaller than predicted by Eq(1).

This discrepancy might be partly associated with the substantially different mechanical properties between Si and SiO2 layers as mentioned before.

It may also be caused by the fact that Eq(1) was derived for the case of blister that is much larger than the wafer thickness t, whereas the actually observed onset blisters usually have the size in the order of the layer thickness t.

It is worth mentioning that the predominant growth of platelets in a closed form before the sudden pop up of surface blisters allows to preanneal the wafers to a time shorter than the onset time at the given preannealing temperature while keeping the wafer suface still good enough for subsequent wafer bonding.

this way, the preannealed wafers can still be bonded and split even at a temperature lower than the preannealed temperature which is desirable for layer transfer between dissimilar materials with grossly different thermal expansion coefficients.

In conclusion, we have found that the formation of surface blisters from hydrogen-containing platelets in hydrogen-implanted silicon consists of a sudden pop-up step at the onset or blistering time rather than a gradual growth.

The blistering time corresponds to the annealing time at which the platelets have reached a critical size through lateral propagation during thermal annealing.

At this point, the hydrogen pressure inside the platelets is sufficiently high to open up the platelets.

The critical size of the onset blisters is found to increase with increasing hydrogen implantation depth or top layer thickness.

The suthors are grateful to Shin-Etsu-Handorai Company and to the German Federeal Ministry of Science and Technology under Contract No.13N6758 for financial support.

Icon 氏名/ニックネーム
Icon 職業/会社名/学校名
Icon 役職/学年
Icon 誕生日
Icon 性別
Icon 出身地/出身校
Icon 住まい
Icon 趣味
Icon 特技
Icon 好きなもの、嫌いなもの
ゲストブックを読む ゲストブックに書く 日記帳 ランキング



2001年12月16日 17時31分19秒

GeoCities Japan

メイン / コミュニティ / アベニュー / E-List / 街角広場
インフォメーションセンター / 検索 / ヘルプ / ガイドライン