# 0.13 Maximum likelihood estimation

 Page 1 / 1

In the last lecture we derived a risk (MSE) bound for regression problems; i.e., select an $f\in \mathcal{F}$ so that $E\left[{\left(f\left(X\right)-Y\right)}^{2}\right]-E\left[{\left({f}^{*}\left(X\right)-Y\right)}^{2}\right]$ is small, where ${f}^{*}\left(x\right)=E\left[Y|X=x\right]$ . The result is summarized below.

Theorem

## Complexity regularization with squared error loss

Let $\mathcal{X}={\mathbb{R}}^{d}$ , $\mathcal{Y}=\left[-b/2,b/2\right]$ , ${\left\{{X}_{i},{Y}_{i}\right\}}_{i=1}^{n}$ iid, ${P}_{XY}$ unknown, $\mathcal{F}$ = {collection of candidate functions},

$f:{\mathbb{R}}^{d}\to \mathcal{Y},\phantom{\rule{4pt}{0ex}}\phantom{\rule{4pt}{0ex}}\phantom{\rule{4pt}{0ex}}R\left(f\right)=E\left[{\left(f\left(X\right)-Y\right)}^{2}\right].$

Let $c\left(f\right)$ , $f\in \mathcal{F}$ , be positive numbers satisfying ${\sum }_{f\in \mathcal{F}}{2}^{-c\left(f\right)}\le 1$ , and select a function from $\mathcal{F}$ according to

${\stackrel{^}{f}}_{n}=argmin\left\{{\stackrel{^}{R}}_{n},\left(f\right),+,\frac{1}{ϵ},\phantom{\rule{0.166667em}{0ex}},\frac{c\left(f\right)log2}{n}\right\},$

with $ϵ\le \frac{3}{5{b}^{2}}$ and ${\stackrel{^}{R}}_{n}\left(f\right)=\frac{1}{n}{\sum }_{i=1}^{n}{\left(f\left({X}_{i}\right)-{Y}_{i}\right)}^{2}$ . Then,

$E\left[R\left({\stackrel{^}{f}}_{n}\right)\right]-R\left({f}^{*}\right)\phantom{\rule{4pt}{0ex}}\le \phantom{\rule{4pt}{0ex}}\left(\frac{1+\alpha }{1-\alpha }\right)\phantom{\rule{0.166667em}{0ex}}\underset{f\in \mathcal{F}}{min}\left\{R,\left(f\right),-,R,\left({f}^{*}\right),+,\frac{1}{ϵ},\phantom{\rule{0.166667em}{0ex}},\frac{c\left(f\right)log2}{n}\right\}+O\left({n}^{-1}\right)$

where $\alpha =\frac{ϵ{b}^{2}}{1-2{b}^{2}ϵ/3}$ .

## Maximum likelihood estimation

The focus of this lecture is to consider another approach to learning based on maximum likelihood estimation. Consider the classical signalplus noise model:

${Y}_{i}=f\left(\frac{i}{n}\right)+{W}_{i},i=1,\cdots ,n$

where ${W}_{i}$ are iid zero-mean noises. Furthermore, assume that ${W}_{i}\sim P\left(w\right)$ for some known density $P\left(w\right)$ . Then

${Y}_{i}\sim P\left(y,-,f,\left(\frac{i}{n}\right)\right)\equiv {P}_{{f}_{i}}\left(y\right)$

since ${Y}_{i}-f\left(\frac{i}{n}\right)={W}_{i}$ .

A very common and useful loss function to consider is

${\stackrel{^}{R}}_{n}\left(f\right)=\frac{1}{n}\sum _{i=1}^{n}\left(-log{P}_{{f}_{i}}\left({Y}_{i}\right)\right).$

Minimizing ${\stackrel{^}{R}}_{n}$ with respect to $f$ is equivalent to maximizing

$\frac{1}{n}\sum _{i=1}^{n}log{P}_{{f}_{i}}\left({Y}_{i}\right)$

or

$\prod _{i=1}^{n}{P}_{{f}_{i}}\left({Y}_{i}\right).$

Thus, using the negative log-likelihood as a loss function leads to maximum likelihood estimation. If the ${W}_{i}$ are iid zero-mean Gaussian r.v.s then this is just the squared error loss weconsidered last time. If the ${W}_{i}$ are Laplacian distributed e.g. $P\left(w\right)\propto {e}^{-|w|}$ , then we obtain the absolute error, or ${L}_{1}$ , loss function. We can also handle non-additive models such as thePoisson model

${Y}_{i}\sim P\left(y|f,\left(i,/,n\right)\right)={e}^{-f\left(i/n\right)}\frac{{\left[f\left(i/n\right)\right]}^{y}}{y!}.$

In this case

$-logP\left({Y}_{i},|f,\left(i,/,n\right)\right)\phantom{\rule{4pt}{0ex}}=\phantom{\rule{4pt}{0ex}}f\left(i,/,n\right)-{Y}_{i}log\left(f,\left(i,/,n\right)\right)+\mathrm{constant}$

which is a very different loss function, but quite appropriate for many imaging problems.

Before we investigate maximum likelihood estimation for model selection, let's review some of the basic concepts. Let $\Theta$ denote a parameter space (e.g., $\Theta =R$ ), and assume we have observations

${Y}_{i}\stackrel{iid}{\sim }{P}_{{\theta }^{*}}\left(y\right),\phantom{\rule{4pt}{0ex}}\phantom{\rule{4pt}{0ex}}i=1,\cdots ,n$

where ${\theta }^{*}\in \Theta$ is a parameter determining the density of the { ${Y}_{i}$ }. The ML estimator of ${\theta }^{*}$ is

$\begin{array}{ccc}\hfill {\stackrel{^}{\theta }}_{n}& =& arg\underset{\theta \in \Theta }{max}\prod _{i=1}^{n}{P}_{\theta }\left({Y}_{i}\right)\hfill \\ & =& arg\underset{\theta \in \Theta }{max}\sum _{i=1}^{n}log{P}_{\theta }\left({Y}_{i}\right)\hfill \\ & =& arg\underset{\theta \in \Theta }{min}\sum _{i=1}^{n}-log{P}_{\theta }\left({Y}_{i}\right).\hfill \end{array}$

$\stackrel{^}{\theta }$ maximizes the expected log-likelihood. To see this, let's compare the expected log-likelihood of ${\theta }^{*}$ with any other $\theta \in \Theta$ .

$\begin{array}{ccc}\hfill E\left[log{P}_{{\theta }^{*}}\left(Y\right)-log{P}_{\theta }\left(Y\right)\right]& =& E\left[log,\frac{{P}_{{\theta }^{*}}\left(Y\right)}{{P}_{\theta }\left(Y\right)}\right]\hfill \\ & =& \int log\frac{{P}_{{\theta }^{*}}\left(y\right)}{{P}_{\theta }\left(y\right)}{P}_{{\theta }^{*}}\left(y\right)dy\hfill \\ & =& K\left({P}_{\theta },{P}_{{\theta }^{*}}\right)\phantom{\rule{1.em}{0ex}}\text{the}\phantom{\rule{4.pt}{0ex}}\text{KL}\phantom{\rule{4.pt}{0ex}}\text{divergence}\hfill \\ & \ge & 0\phantom{\rule{1.em}{0ex}}\text{with}\phantom{\rule{4.pt}{0ex}}\text{equality}\phantom{\rule{4.pt}{0ex}}\text{iff}\phantom{\rule{0.166667em}{0ex}}{P}_{{\theta }^{*}}={P}_{\theta }.\hfill \end{array}$

Why?

$\begin{array}{ccc}\hfill -E\left[log,\frac{{P}_{{\theta }^{*}}\left(y\right)}{{P}_{\theta }\left(y\right)}\right]& =& E\left[log,\frac{{P}_{\theta }\left(y\right)}{{P}_{{\theta }^{*}}\left(y\right)}\right]\hfill \\ & \le & logE\left[\frac{{P}_{\theta }\left(y\right)}{{P}_{{\theta }^{*}}\left(y\right)}\right]\hfill \\ & =& log\int {P}_{\theta }\left(y\right)dy=0\hfill \\ & ⇒& K\left({P}_{\theta },{P}_{{\theta }^{*}}\right)\ge 0\hfill \end{array}$

On the other hand, since ${\stackrel{^}{\theta }}_{n}$ maximizes the likelihood over $\theta \in \Theta$ , we have

$\sum _{i=1}^{n}log\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{{\stackrel{^}{\theta }}_{n}}\left({Y}_{i}\right)}=\sum _{i=1}^{n}log{P}_{{\theta }^{*}}\left({Y}_{i}\right)-log{P}_{{\stackrel{^}{\theta }}_{n}}\left({Y}_{i}\right)\le 0.$

Therefore,

$\frac{1}{n}\sum _{i=1}^{n}log\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{{\stackrel{^}{\theta }}_{n}}\left({Y}_{i}\right)}-K\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)+K\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\le 0$

or re-arranging

$K\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\le \left|\frac{1}{n},\sum _{i=1}^{n},log,\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{{\stackrel{^}{\theta }}_{n}}\left({Y}_{i}\right)},-,K,\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\right|.$

Notice that the quantity

$\frac{1}{n}\sum _{i=1}^{n}log\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{\theta }\left({Y}_{i}\right)}$

is an empirical average whose mean is $K\left({P}_{\theta },{P}_{{\theta }^{*}}\right)$ . By the law of large numbers, for each $\theta \in \Theta$ ,

$\left|\frac{1}{n},\sum _{i=1}^{n},log,\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{\theta }\left({Y}_{i}\right)},-,K,\left({P}_{\theta },{P}_{{\theta }^{*}}\right)\right|\stackrel{a.s.}{\to }0.$

If this also holds for the sequence $\left\{{\stackrel{^}{\theta }}_{n}\right\}$ , then we have

$K\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\le \left|\frac{1}{n},\sum ,log,\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{{\stackrel{^}{\theta }}_{n}}\left({Y}_{i}\right)},-,K,\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\right|\to 0\phantom{\rule{0.166667em}{0ex}}\mathrm{as}\phantom{\rule{0.166667em}{0ex}}n\to \infty$

which implies that

${P}_{{\stackrel{^}{\theta }}_{n}}\to {P}_{{\theta }^{*}}$

which often implies that

${\stackrel{^}{\theta }}_{n}\to {\theta }^{*}$

in some appropriate sense (e.g., point-wise or in norm).

## Gaussian distributions

${P}_{{\theta }^{*}}\left(y\right)=\frac{1}{\sqrt{\pi }}{e}^{-{\left(y-{\theta }^{*}\right)}^{2}}$
$\Theta =\mathbb{R},\phantom{\rule{1.em}{0ex}}{\left\{{Y}_{i}\right\}}_{i=1}^{n}\stackrel{iid}{\sim }{P}_{{\theta }^{*}}\left(y\right)$
$\begin{array}{ccc}\hfill K\left({P}_{\theta },{P}_{{\theta }^{*}}\right)& =& \int log\frac{{P}_{{\theta }^{*}}\left(y\right)}{{P}_{\theta }\left(y\right)}{P}_{{\theta }^{*}}\left(y\right)dy\hfill \\ & =& \int \left[{\left(y-\theta \right)}^{2}-{\left(y-{\theta }^{*}\right)}^{2}\right]{P}_{{\theta }^{*}}\left(y\right)dy\hfill \\ & =& {E}_{{\theta }^{*}}\left[{\left(y-\theta \right)}^{2}\right]-{E}_{{\theta }^{*}}\left[{\left(y-{\theta }^{*}\right)}^{2}\right]\hfill \\ & =& {E}_{{\theta }^{*}}\left[{Y}^{2}-2Y\theta +{\theta }^{2}\right]-1/2\hfill \\ & =& {\left({\theta }^{*}\right)}^{2}+1/2-2{\theta }^{*}\theta +{\theta }^{2}-1/2\hfill \\ & =& {\left({\theta }^{*}-\theta \right)}^{2}\hfill \end{array}$
$⇒{\theta }^{*}\phantom{\rule{0.166667em}{0ex}}\text{maximizes}\phantom{\rule{0.166667em}{0ex}}E\left[log{P}_{\theta }\left(Y\right)\right]\phantom{\rule{0.166667em}{0ex}}\text{wrt}\phantom{\rule{0.166667em}{0ex}}\theta \in \Theta$
$\begin{array}{ccc}\hfill {\stackrel{^}{\theta }}_{n}& =& arg\underset{\theta }{max}\left\{-\sum {\left({Y}_{i}-\theta \right)}^{2}\right\}\hfill \\ & =& arg\underset{\theta }{min}\left\{\sum {\left({Y}_{i}-\theta \right)}^{2}\right\}\hfill \\ & =& \frac{1}{n}\sum _{i=1}^{n}{Y}_{i}\hfill \end{array}$

## Hellinger distance

The KL divergence is not a distance function.

$K\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)\ne K\left({P}_{{\theta }_{2}},{P}_{{\theta }_{1}}\right)$

Therefore, it is often more convenient to work with the Hellinger metric,

$H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)={\left(\int ,{\left({P}_{{\theta }_{1}}^{\frac{1}{2}},-,{P}_{{\theta }_{2}}^{\frac{1}{2}}\right)}^{2},d,y\right)}^{\frac{1}{2}}.$

The Hellinger metric is symmetric, non-negative and

$H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)=H\left({P}_{{\theta }_{2}},{P}_{{\theta }_{1}}\right)$

and therefore it is a distance measure. Furthermore, the squared Hellinger distance lower bounds the KL divergence, so convergence in KL divergence implies convergence of the Hellinger distance.

## Proposition 1

${H}^{2}\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)\le K\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)$

## Proof:

$\begin{array}{ccc}\hfill H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)& =& \int {\left(\sqrt{{P}_{{\theta }_{1}}\left(y\right)},-,\sqrt{{P}_{{\theta }_{2}}\left(y\right)}\right)}^{2}dy\hfill \\ & =& \int {P}_{{\theta }_{1}}\left(y\right)dy+\int {P}_{{\theta }_{2}}\left(y\right)dy-2\int \sqrt{{P}_{{\theta }_{1}}\left(y\right)}\sqrt{{P}_{{\theta }_{2}}\left(y\right)}dy\hfill \\ & =& 2-2\int \sqrt{{P}_{{\theta }_{1}}\left(y\right)}\sqrt{{P}_{{\theta }_{2}}\left(y\right)}dy,\phantom{\rule{1.em}{0ex}}\text{since}\phantom{\rule{0.166667em}{0ex}}\int {P}_{\theta }\left(y\right)dy=1\phantom{\rule{0.166667em}{0ex}}\forall \theta \hfill \\ & =& 2\left(1,-,{E}_{{\theta }_{2}},\left[\sqrt{{P}_{{\theta }_{1}}\left(Y\right)/{P}_{{\theta }_{2}}\left(Y\right)}\right]\right)\hfill \\ & \le & 2log\left({E}_{{\theta }_{2}},\left[\sqrt{{P}_{{\theta }_{2}}\left(Y\right)/{P}_{{\theta }_{1}}\left(Y\right)}\right]\right),\phantom{\rule{1.em}{0ex}}\text{since}\phantom{\rule{0.166667em}{0ex}}1-x\le -logx\hfill \\ & \le & 2{E}_{{\theta }_{2}}\left[log,\sqrt{{P}_{{\theta }_{2}}\left(Y\right)/{P}_{{\theta }_{1}}\left(Y\right)}\right],\phantom{\rule{1.em}{0ex}}\text{by}\phantom{\rule{4.pt}{0ex}}\text{Jensen's}\phantom{\rule{4.pt}{0ex}}\text{inequality}\hfill \\ & =& {E}_{{\theta }_{2}}\left[log,\left(,{P}_{{\theta }_{2}},\left(Y\right),/,{P}_{{\theta }_{1}},\left(Y\right),\right)\right]\phantom{\rule{4pt}{0ex}}\equiv \phantom{\rule{4pt}{0ex}}K\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)\hfill \end{array}$

Note that in the proof we also showed that

$H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)=2\left(1,-,\int ,\sqrt{{P}_{{\theta }_{1}}\left(y\right)},\sqrt{{P}_{{\theta }_{2}}\left(y\right)},d,y\right)$

and using the fact $logx\le x-1$ again, we have

$H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)\le -2log\left(\int ,\sqrt{{P}_{{\theta }_{1}}\left(y\right)},\sqrt{{P}_{{\theta }_{2}}\left(y\right)},d,y\right).$

The quantity inside the log is called the affinity between ${P}_{{\theta }_{1}}$ and ${P}_{{\theta }_{2}}$ :

$A\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)=\int \sqrt{{P}_{{\theta }_{1}}\left(y\right)}\sqrt{{P}_{{\theta }_{2}}\left(y\right)}dy.$

This is another measure of closeness between ${P}_{{\theta }_{1}}$ and ${P}_{{\theta }_{2}}$ .

## Gaussian distributions

${P}_{\theta }\left(y\right)=\frac{1}{\pi }{e}^{-{\left(y-\theta \right)}^{2}}$
$\begin{array}{ccc}& & -2log\int \sqrt{{P}_{{\theta }_{1}}\left(y\right)}\sqrt{{P}_{{\theta }_{2}}\left(y\right)}dy\hfill \\ & =& -2log\int {\left(\frac{1}{\sqrt{\pi }},{e}^{-{\left(y-{\theta }_{1}\right)}^{2}}\right)}^{\frac{1}{2}}{\left(\frac{1}{\sqrt{\pi }},{e}^{-{\left(y-{\theta }_{2}\right)}^{2}}\right)}^{\frac{1}{2}}dy\hfill \\ & =& -2log\left(\int ,\frac{1}{\sqrt{\pi }},{e}^{-\left[\frac{{\left(y-{\theta }_{1}\right)}^{2}}{2},+,\frac{{\left(y-{\theta }_{2}\right)}^{2}}{2}\right]},d,y\right)\hfill \\ & =& -2log\left(\int ,\frac{1}{\sqrt{\pi }},{e}^{-\left[{\left(y,-,\left(,\frac{{\theta }_{1}+{\theta }_{2}}{2},\right)\right)}^{2},+,{\left(\frac{{\theta }_{1}-{\theta }_{2}}{2}\right)}^{2}\right]},d,y\right)\hfill \\ & =& -2log{e}^{-{\left(\frac{{\theta }_{1}-{\theta }_{2}}{2}\right)}^{2}}\hfill \\ & =& \frac{1}{2}{\left({\theta }_{1}-{\theta }_{2}\right)}^{2}\hfill \end{array}$
$\begin{array}{ccc}& ⇒& -2logA\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)=\frac{1}{2}{\left({\theta }_{1}-{\theta }_{2}\right)}^{2}\phantom{\rule{1.em}{0ex}}\text{for}\phantom{\rule{4.pt}{0ex}}\text{Gaussian}\phantom{\rule{4.pt}{0ex}}\text{distributions}\hfill \\ & ⇒& H\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)\le \frac{1}{2}{\left({\theta }_{1}-{\theta }_{2}\right)}^{2}\phantom{\rule{1.em}{0ex}}\text{for}\phantom{\rule{4.pt}{0ex}}\text{Gaussian}.\hfill \end{array}$

## Poisson distributions

If ${P}_{\theta }\left(y\right)={e}^{-\theta }\frac{{\theta }^{y}}{y!},\theta \ge 0$ , then

$-2logA\left({P}_{{\theta }_{1}},{P}_{{\theta }_{2}}\right)={\left(\sqrt{{\theta }_{1}}-\sqrt{{\theta }_{2}}\right)}^{2}.$

## Summary

${Y}_{i}\stackrel{iid}{\sim }{P}_{{\theta }^{*}}$
1. Maximum likelihood estimator maximizes the empirical average
$\frac{1}{n}\sum _{i=1}^{n}log{P}_{\theta }\left({Y}_{i}\right)$
(our empirical risk is negative log-likelihood)
2. ${\theta }^{*}$ maximizes the expectation
$E\left[\frac{1}{n},\sum _{i=1}^{n},log,{P}_{\theta },\left({Y}_{i}\right)\right]$
(the risk is the expected negative log-likelihood)
3. $\frac{1}{n}\sum _{i=1}^{n}log{P}_{\theta }\left({Y}_{i}\right)\stackrel{a.s.}{\to }E\left[\frac{1}{n},\sum _{i=1}^{n},log,{P}_{\theta },\left({Y}_{i}\right)\right]$
so we expect some sort of concentration of measure.
4. In particular, since
$\frac{1}{n}\sum _{i=1}^{n}log\frac{{P}_{{\theta }^{*}}\left({Y}_{i}\right)}{{P}_{\theta }\left({Y}_{i}\right)}\stackrel{a.s.}{\to }K\left({P}_{\theta },{P}_{{\theta }^{*}}\right)$
we might expect that $K\left({P}_{{\stackrel{^}{\theta }}_{n}},{P}_{{\theta }^{*}}\right)\to 0$ for the sequence of estimates ${\left\{{P}_{{\stackrel{^}{\theta }}_{n}}\right\}}_{n=1}^{\infty }$ .

So, the point is that maximum likelihood estimator is just a special case of a loss function in learning. Due to its special structure, weare naturally led to consider KL divergences, Hellinger distances, and Affinities.

Is there any normative that regulates the use of silver nanoparticles?
what king of growth are you checking .?
Renato
What fields keep nano created devices from performing or assimulating ? Magnetic fields ? Are do they assimilate ?
why we need to study biomolecules, molecular biology in nanotechnology?
?
Kyle
yes I'm doing my masters in nanotechnology, we are being studying all these domains as well..
why?
what school?
Kyle
biomolecules are e building blocks of every organics and inorganic materials.
Joe
anyone know any internet site where one can find nanotechnology papers?
research.net
kanaga
sciencedirect big data base
Ernesto
Introduction about quantum dots in nanotechnology
what does nano mean?
nano basically means 10^(-9). nanometer is a unit to measure length.
Bharti
do you think it's worthwhile in the long term to study the effects and possibilities of nanotechnology on viral treatment?
absolutely yes
Daniel
how to know photocatalytic properties of tio2 nanoparticles...what to do now
it is a goid question and i want to know the answer as well
Maciej
Abigail
for teaching engĺish at school how nano technology help us
Anassong
Do somebody tell me a best nano engineering book for beginners?
there is no specific books for beginners but there is book called principle of nanotechnology
NANO
what is fullerene does it is used to make bukky balls
are you nano engineer ?
s.
fullerene is a bucky ball aka Carbon 60 molecule. It was name by the architect Fuller. He design the geodesic dome. it resembles a soccer ball.
Tarell
what is the actual application of fullerenes nowadays?
Damian
That is a great question Damian. best way to answer that question is to Google it. there are hundreds of applications for buck minister fullerenes, from medical to aerospace. you can also find plenty of research papers that will give you great detail on the potential applications of fullerenes.
Tarell
what is the Synthesis, properties,and applications of carbon nano chemistry
Mostly, they use nano carbon for electronics and for materials to be strengthened.
Virgil
is Bucky paper clear?
CYNTHIA
carbon nanotubes has various application in fuel cells membrane, current research on cancer drug,and in electronics MEMS and NEMS etc
NANO
so some one know about replacing silicon atom with phosphorous in semiconductors device?
Yeah, it is a pain to say the least. You basically have to heat the substarte up to around 1000 degrees celcius then pass phosphene gas over top of it, which is explosive and toxic by the way, under very low pressure.
Harper
Do you know which machine is used to that process?
s.
how to fabricate graphene ink ?
for screen printed electrodes ?
SUYASH
What is lattice structure?
of graphene you mean?
Ebrahim
or in general
Ebrahim
in general
s.
Graphene has a hexagonal structure
tahir
On having this app for quite a bit time, Haven't realised there's a chat room in it.
Cied
what is biological synthesis of nanoparticles
how did you get the value of 2000N.What calculations are needed to arrive at it
Privacy Information Security Software Version 1.1a
Good
Got questions? Join the online conversation and get instant answers!