Diamond Chemical Vapor Deposition
BULLDAWG
There seems to be some interest in how to initiate nucleation on the
substrate.
To put things into perspective diamond growth, regardless of the method
used to deposit the stuff generally follows this pattern:
1. induction, where nothing much happens
2. nucleation where seed diamond particles form on the surface
3. growth, where diamond is deposited.
I have a reference, "Towards a general concept of diamond deposition", Bachman
et al, Diamond and Related Materials, 1 (1991) 1-12. This a pretty good
reference and I will refer to it as simply Bachman et al.
Anyway, some people wrote me and asked if I did some special things to promote
nucleation. The answer is not really.
Generally I just scratch the Silicon wafer with 0.5 to 1 um diamond grit seeded
into acetone in an ultrasound machine.
But what about scratching? I have gobs of references that say you must "damage"
the surface somehow. Obviously scratching the surface causes damage. The most
prevalent means of damging the surface are:
1. scratching with some material (and the material does not need to be diamond).
2. ion beam implantation (fast atom bombardment or FAB)
3. photolithography (pattern etching).
4. other.
And of course, for some applications one wants to be able to selectivaly
deposit diamond. The only means of inhibiting deposition that I have seen
are:
1. ion beam annealing of the surface (silicon typically) so as to produce
amorphous substrate structure.
2. selective scratching.
However, in my work, I have demonstrated (yet to be published) that one does
not need to scratch your substrate to effect full uniform coverage over a
few inches.
So I wonder, what is it about surface damage that seems to enhance nucleation
rates? And why under certain conditions it is possible to lay down diamond film
on undamaged surfaces?
I have read some liturature work where they etched certain patterns on the
substrate surface. Such as cones on the order of microns. They found that
diamond selectivally formed first on the bases. Same thing when they formed
trenches, deposition in the trench not the ridge. Not only was this observed
but the size of the scratches were critical. It seemd that scratches on the
order of 5 um to 15 um were best.
This suggests to me that one needs stagnation of the feed gas on the surface to
effect nucleation.
Ken Pearce
Derrick C. Mancini
<kpe...@magnus.acs.ohio-state.edu> wrote:
> Thanks to all who responded (via e-mail) to my previous article.
> But what about scratching? I have gobs of references that say you
>must "damage" the surface somehow. Obviously scratching the surface
>causes damage. The most prevalent means of damging the surface are:
> 1. scratching with some material (and the material does not need to be
>diamond).
> 2. ion beam implantation (fast atom bombardment or FAB)
> 3. photolithography (pattern etching).
> 4. other.
While the "scratching" compound does not have to be diamond, studies
have show that scratching with diamond has a much higher nucleation rate
than scratching with BN than scratching with other materials than making
the surface rough or etching. Most likely, the diamond grits provide
true "pre-nucleated" sites for diamond growth, the BN provides nearly
lattice-matched sites for nucleation, and thereafter the defects are
likely nucleations sites, as they are for many nucleation phenomena,
since homogeneous nucleation is fairly rare.
> So I wonder, what is it about surface damage that seems to enhance nucleation
> rates? And why under certain conditions it is possible to lay down
>diamond film on undamaged surfaces?
see above
> I have read some liturature work where they etched certain patterns on the
> substrate surface. Such as cones on the order of microns. They found that
> diamond selectivally formed first on the bases. Same thing when they formed
> trenches, deposition in the trench not the ridge. Not only was this observed
> but the size of the scratches were critical. It seemd that scratches on the
> order of 5 um to 15 um were best.
> This suggests to me that one needs stagnation of the feed gas on the
>surface to effect nucleation.
Another explanation is that you have to form a diamond nanocluster to
provide a nucleation site. Since the action of the hydrogen in the reaction
chamber is to etch away any graphitic carbon, and any carbon that is not sp3
hybridized (and hence has 4 bonding partners) will thus be removed, a
corner site would provide a greater chance of protecting a carbon in sp3
state by providing more binding sites on the substrate. While access of
hydrogren is reduced in the corners, so is hydrocarbon. So I doubt that
it is just access of the reacting gas that allows growth. But what
carbon might bind in a corner can protect more of its bonds. Once the
nanocluster is formed, there you go.
Derrick C. Mancini
Fysiska Institutionen - Uppsala Universitet
Box 530 S751 21 Uppsala Sverige
man...@fysik.uu.se
---------------------------------------------
"Are we having fun yet?" - Zippy
BULLDAWG
>In article <1993Feb23....@magnus.acs.ohio-state.edu>, BULLDAWG
><kpe...@magnus.acs.ohio-state.edu> wrote:
>While the "scratching" compound does not have to be diamond, studies
>have show that scratching with diamond has a much higher nucleation rate
>than scratching with BN than scratching with other materials than making
>the surface rough or etching.
I don't really think that is correct. Recent studies have shown that
the mechanics of surface damgae are largely irrelevant. You can prepare
a sample with scratching with a variety of techniques and with various
detection methods (raman, xps, ir, sem) one can be assured that there are
no "seed particles" left behind, which I think is what you are suggesting.
>> I have read some literature work where they etched certain patterns on the
>> substrate surface. Such as cones on the order of microns. They found that
>> diamond selectivally formed first on the bases. Same thing when they formed
>> trenches, deposition in the trench not the ridge. Not only was this observed
>> but the size of the scratches were critical. It seemd that scratches on the
>> order of 5 um to 15 um were best.
>> This suggests to me that one needs stagnation of the feed gas on the
>>surface to effect nucleation.
>Another explanation is that you have to form a diamond nanocluster to
>provide a nucleation site.
I really don't think that one can consider a micron size structure to have
atomic scale corners. I really think that it is a gas dynamics on the
macro-atomic scale that is responsible.
[Moderator's note: please trim quoted articles as much as possible
to keep the new/old info ratio high. I am actually UNABLE to post articles
for which this ratio is < 1/2, and had to trim this one down quite a bit.]
j...@rti.rti.org
>In article <mancini-25...@usx221.fysik.uu.se> man...@fysik.uu.se
>(Derrick C. Mancini) writes:
>>While the "scratching" compound does not have to be diamond, studies
>>have show that scratching with diamond has a much higher nucleation rate
>>than scratching with BN than scratching with other materials than making
>>the surface rough or etching.
>I don't really think that is correct. Recent studies have shown that
>the mechanics of surface damgae are largely irrelevant. You can prepare
>a sample with scratching with a variety of techniques and with various
>detection methods (raman, xps, ir, sem) one can be assured that there are
>no "seed particles" left behind, which I think is what you are suggesting.
It may be of interest to note that S. Iijima of NEC did a nice piece of
work with regards to this matter that was published a while (~ 1 year)
ago in J. Materials Research. He used plan-view TEM samples of diamond
scratched Si wafers and saw that there were tiny diamond crystals left
after scatching and showed via iterative growth/TEM/growth/etc. on the
same area that the little diamond particles were where the diamond
growth first occurred. There is, however, more to this issue, as
several methods of treating sample surfaces with a non-diamond carbon
source will also enhance the diamond nucleation kinetics.
--
John Posthill
Derrick C. Mancini
kpe...@magnus.acs.ohio-state.edu (BULLDAWG) wrote:
> I don't really think that is correct. Recent studies have shown that
> the mechanics of surface damgae are largely irrelevant. You can prepare
> a sample with scratching with a variety of techniques and with various
> detection methods (raman, xps, ir, sem) one can be assured that there are
> no "seed particles" left behind, which I think is what you are suggesting.
Yes, that is my suggestion. But I am not aware of many studies that
suggest "mechanics of surface damage" are "irrelevant". There are those
that suggest that film growth can occur under a wide variety of damage
mechanisms. Unfortunately, few studies control all variables well, which
is necessary to determine the true nature of nucleation. As another poster
has pointed out, there are studies that show residual seed particles are
the site of nucleation for some class of films. Also, it is this "pre-
seeding method" that is used by others to grow large, quasi-single
crystal diamond on silicon (see recent Sci. Amer. review). However, I
agree that growth can occur due to damage alone. Could you point out
those studies which show NO relevance of damage mechanics (as opposed
to showing no evidence one way or the other)?
> >Another explanation is that you have to form a diamond nanocluster to
> >provide a nucleation site.
> I really don't think that one can consider a micron size structure to have
> atomic scale corners. I really think that it is a gas dynamics on the
> macro-atomic scale that is responsible.
I agree, a micron size structure would not apply. That is why I used
the term nanocluster, usually 1-10 nanometers in size (10-100 angstroms),
at which scale the surface features may well show atomic-scale effects.
This is their salient feature, they are just large enough to show long-
range order in their structure.
(BTW, I don't know what macro-atomic scale is - could you specify?)
Again, what bothers me about the gas dynamics explanation is that it
it doesn't suggest why the decrease access of carbon species doesn't
reduce nucleation as much as decrease access of hydrogren reduces
etching. That doesn't rule it out, just doesn't support it, I think.
BULLDAWG
(Derrick C. Mancini) writes:
>In article <1993Feb26.1...@magnus.acs.ohio-state.edu>,
>kpe...@magnus.acs.ohio-state.edu (BULLDAWG) wrote:
>> You can prepare
>> a sample with scratching with a variety of techniques and with various
>> detection methods (raman, xps, ir, sem) one can be assured that there are
>> no "seed particles" left behind, which I think is what you are suggesting.
>Yes, that is my suggestion.
[deleted stuff]
>However, I
>agree that growth can occur due to damage alone. Could you point out
>those studies which show NO relevance of damage mechanics (as opposed
>to showing no evidence one way or the other)?
OK, I have references, which I will post shortly that show that the mechanism
of damage, as long as the "damage" is defined as disruption of bonds, not of
creating amorphous "smooth" structures, is largely irrelevant. Perhaps my
words were too exclusive.
>> >Another explanation is that you have to form a diamond nanocluster to
>> >provide a nucleation site.
>> I really don't think that one can consider a micron size structure to have
>> atomic scale corners. I really think that it is a gas dynamics on the
>> macro-atomic scale that is responsible.
>I agree, a micron size structure would not apply. That is why I used
>the term nanocluster, usually 1-10 nanometers in size (10-100 angstroms),
>at which scale the surface features may well show atomic-scale effects.
>This is their salient feature, they are just large enough to show long-
>range order in their structure.
>(BTW, I don't know what macro-atomic scale is - could you specify?)
Ummmm, got me again for being a little sloppy. On the scale of 0.1 microns
or larger is what I meant.
>Again, what bothers me about the gas dynamics explanation is that it
>it doesn't suggest why the decrease access of carbon species doesn't
>reduce nucleation as much as decrease access of hydrogren reduces
>etching. That doesn't rule it out, just doesn't support it, I think.
I think it has not so much to due with lack of *access* because these
structures are many times larger than atoms (structure size = 12
microns). The suggested mechanism is that there are either (1)
differences in temperature from ridge to valley (and it has been shown
that growth / nucleation density is strongly effected by temperature) or
(2) from stagnation of species in the valleys.
As far as evidence that one *does not* need diamond-like seed particles, I
have in my lab, grown films on (1) non-scratched polished Si wafers and (2)
grown films on wafers damaged by Aluminum oxide grit (sand-blasted)