The heavily boron-doped single crystal produced by CSMH

In brief, heavily boron-doped single crystal SiNMs are first formed on an SOI substrate and then released by selective etching-away of bur- ied oxide. The released  SiNMs  are transferred to  diamond substrate via the  stamp-assisted  transfer  printing  method.20 The diamond plate bearing the SiNMs is annealed via RTA to first form a stronger bonding and then to induce boron diffu- sion  from  Si  into  diamond.  The  doping  mechanism  is described later in the text.

The thermal diffusion doping method has a comparative advantage  over  ion  implantation  in  that  lattice  structural damages   are   not   introduced   during   thermal   diffusion. Therefore,  the  high  temperature  recrystallization  process needed for post ion implantation is no longer necessary and graphitization can be readily avoided. It is also expected that higher crystal quality can be obtained using the thermal dif- fusion method as opposed to the ion implantation method af- ter  finishing  the  doping  process.  By  using  the  transfer printing  method,  clean  interfaces  are  ensured  and  impor- tantly, selective doping (to be seen later), via deterministic transfer printing of SiNMs of different sizes to the selective areas on the diamond surface, is made easy and precise.21 The selective doping enabled by selective transfer printing, while applicable to any size and shape of diamond plates (in contrast to direct wafer bonding), leads to a planar doped structure  that  can  facilitate  device  implementations.  Figs. 1(b) and 1(c) show images of natural diamond before and af- ter SiNM bonding.

Diamond crystal structures before and after boron diffu- sion doping are first characterized and then compared. Before performing the characterizations, the SiNM is removed using potassium hydroxide (KOH) after completion of the RTA pro- cess. No graphitization removal procedures were applied to the diamond surface. After finishing the Si removal procedure, we are unable to identify any remaining Si or SiC materials in diamond  from   a  Raman   spectrum  taken  in  the  range   of 400– 1200cm-1  (see Fig. S7).43  The X-ray theta-2theta  scan results of the boron diffused diamond are shown in Fig. 1(d). The (220) peak of the nSCD, before and after the boron diffu- sion process, appeared at 75.825o  in both cases. Of more im- portance, the full width  at half maximum  (FWHM)  of the diamond (220) diffraction peak shows no measurable changes after finishing the diffusion process. The small FWHM value of 0.025o   indicates  the high  single  crystallinity  of the  dia- mond.22,23  The Raman spectra of the nSCD before and after diffusion doping are shown in Fig. 1(e) for comparison. The zoomed-in view of the relevant wave number range of Fig. 1(e) is  shown  in Fig. 1(f).  The  sp3  bonding  in the  sample before and after the thermal diffusion process is clearly indi- cated by the TO phonon peak at 1332cm-1. The FWHM of

the Raman peak became slightly wider and shifted after diffu- sion (from 3.9cm-1  to 4.6cm-1  and 0.4cm-1  of blue-shift). Such a small change could be attributed to the change of the existing crystal imperfection in the nSCD or small stress dur- ing the process. Our nSCD before processing does not have as small FWHM value (2–3cm-1)24–26  in comparison with some of the others that are reported in literature, indicating the exis- tence of some crystal imperfection. However, the FWHM is smaller than 5cm-1  in both cases, further indicating that the sp3 bonding in the diamond remained intact after finishing the boron diffusion process.27,28 It is noted that the FWHM values (13.3cm-1–87.8cm-1)26,29 of Raman spectra in ion implanted (after anneal) SCDs are much larger than 5cm-1.

The X-ray diffraction (XRD) and Raman characteriza- tions indicate that the boron doping method via SiNM bond- ing and thermal diffusion does not induce measurable lattice damage in diamond. Furthermore, as shown in Fig. 1(f) it is noted that the absence of peaks near the wave numbers of 1357cm-1  and  1556cm-1  in the zoomed-in spectra of Fig. 1(e), which are the characteristic indicator of the presence of sp2  bonds,  proves that the  SiNM doping process does not induce detectable graphitization in the diamond bulk or on its  surface.  As  comparison,  Fig.  1(g)  shows  the  Raman spectrum scanned from a reference nSCD sample that has no SiNM bonded but was subject to the identical RTA process. The distinct Raman peak at the wave number of 1556cm—1 in the spectrum clearly indicates the existence of sp2 bonds that are formed on this undoped sample. To further verify the role of SiNM in preventing graphitization on the dia- mond surface, the Raman spectrum taken from the backside of the SiNM bonded diamond (associated with Fig. 1(e) and 1(f)), where no SiNM is bonded, also shows a visible peak at the wave number of 1556cm—1  (see Fig. S3). These results indicate that it is because of using single crystal SiNM as the dopant carrying medium for thermal diffusion that we have successfully avoided graphitization on the diamond surface.

The diffused boron atom concentration in the nSCD is characterized  by  both  secondary  ion  mass  spectroscopy (SIMS) and capacitance-voltage (C—V) measurements. The results are shown in Fig. 2(a) and the fitted curves by the Fick’s law of diffusion are shown in Fig. S4. Fig. 2(a) indi- cates  the  presence  of  boron  at  a  concentration  of  about 1 × 1019 cm—3  at the diamond surface, which is comparable with the level that can be achieved by ion implantation, and gradually   decreased   to   ~2 × 1015 cm—3     at   a   depth   of ~70nm. As a comparison,  ~120cm2/v.s of hole mobility and the doping concentration of ~2 × 1018 cm—3  were meas- ured by using Hall measurements (Accent HL5500 Hall sys- tem). Considering that the annealing time is only 40 min, the doping  depth  achieved  is  encouraging  for  device  applica- tions. The profile obtained by C—V measurements roughly matches that of SIMS in terms of shape and depth consider- ing the limited accuracy of SIMS. The additional experiment under various diffusion time and temperature conditions is under investigation.

Further  characterizations  of  boron-doped  nSCD  were performed  using  Fourier  transform  infrared   spectroscopy (FTIR). Generally, boron inactivation could result from non- substitutional boron sites30  or aggregated substitutional bo- ron sites.31,32 FTIR is an effective method for evaluating the substitutional doping status in a diamond. The FTIR results are shown in Fig. 2(b). Fig. 2(b–i) shows two diamond plates of the same type: one is boron doped (scanning region B) that is realized using the above thermal diffusion method and the  other  is  undoped  (scanning  region  C)  as  a  reference. Figs. 2(b-ii)–(2-iv) show the FTIR mapping results and the scanned spectra. The boron doped diamond shows the char- acteristic absorption peak at 1290cm—1, which clearly indi- cates the electrical activation of boron atoms.33  In contrast, the peak does not appear in the undoped reference diamond. Since substitutional doping, i.e., boron-carbon sp3  bonding formation, is necessary for electrical activation of doped bo- ron atoms, the FTIR and the C—V characterization results prove the substitutional doping of boron atoms in the nSCD. It should be noted that the characteristic absorption peaks associated with boron interstitials and boron interstitial com- plexes in diamond can be observed at 1420, 1530, 1570, and 1910cm—1, but no such peaks appeared in our FTIR spec- tra.34 Moreover, the absence of the three infrared B-B cluster absorption peaks (553, 560, and 570cm—1) indicate that no aggregated  substitutional  boron  sites  were  formed.35    The broader peaks, which appear from 1900 to 2300cm—1  are the inherent two-phonon lines of diamond associated with C—C bonds. They appear in both the boron doped and undoped di- amond samples.

X-ray photoelectron spectroscopy (XPS) was performed on both doped diamond and undoped diamond (as a refer- ence). The binding energies for Si1s, B1s  and C1s  peaks have been  identified  with  constant  pass  energy  of  50eV   and 100meV energy step as shown in Fig. 2(c). The C–Si peak at ~103 eV indicates that a chemical reaction between Si and carbon atoms has occurred at the Si–diamond interface (Fig. 2(c-i). However, the absence of an Si peak in the Raman spectrum (Fig. S7) in the range of 400–1200cm—1  indicates that the C-Si reaction occurred very shallowly (<10nm) at the   diamond   surface.43     Boron   substitution   in   diamond yielded a B1s  peak at  190.6eV corresponding to B–C (Fig. 2(c-ii)) confirming that boron doping was successful. This is consistent  with the  SIMS,  C—V,  and  FTIR  analyses.  Fig. 2(c–iii) shows the XPS spectra for the core C1s  peak in the binding energy region around 280–295eV for the undoped and boron doped nSCD samples. De-convoluted C1s  peaks using the Gaussian/Lorentzian function in the inset of Fig. 2(c-iii) shows a strong sp3  C–C bonding at 285.4eV and a very   small   C–O   and   C = O   bonding   at   286.4eV   and 287.9eV, respectively, indicating that the single crystallinity of nSCD was not degraded by the boron diffusion process.

The XPS results obtained above suggest clues for eluci- dating the  boron  diffusion  doping  mechanisms.  Especially, the C-Si bonding (Fig. 2(c-i) plays an important role for the observed boron diffusion. First-principles density functional theory (DFT) simulations36,37  were performed to understand the diffusion mechanism. Two mechanisms were proposed to yield enhanced boron diffusion through enhanced vacancies in diamond (detailed calculation can be found in the supple- mentary material). The first mechanism involves injection of excess vacancies into the diamond from the SiNM, which has a much larger intrinsic vacancy concentration than diamond as well as excess vacancies from ion implantation into the SiNM. DFT calculations verify that this vacancy injection is much more energetically favorable than the usual mechanism for vacancy formation in diamond (movement of carbon to the diamond surface). Fig. 2(d) shows the cartoon illustration of the proposed injection and diffusion mechanism. Besides the  above  vacancy  injection  into  diamond  from  Si,  excess vacancies can also be  additionally created by formation of SiC at the SiNM and diamond interface, which stabilized the vacancies by almost exactly the required 0.7 eV needed to explain the enhanced boron diffusion rate into diamond. Both mechanisms could play a role and further research is required to elucidate their contributions. In each case, diffused boron atoms from Si are expected to immediately become substitu- tional atoms in diamond if the diamond does not have pre- existing vacancy related  defects  (ideal  situation).  Since  no vacancy defects are additionally generated in diamond during the diffusion process, a high temperature anneal that is neces- sary  for  post-implantation38    then  becomes  unnecessary  in such a doping process.

The boron-doped single-crystal diamond produced by CSMH can achieve doping from low concentration to high concentration. It has realized a uniform and controllable concentration and a customizable boron doping process.CSMH uses the MPCVD method to prepare large-sized and high-quality diamonds,and currently has mature products such as diamond heat sinks, diamond wafers, diamond windows,diamond hetero junction integrated composite substrates,etc.