Bummer!

At that place where I used it, they kept it in the 'fridge in the kitchen. ;-)

Cheers! Rich

Looks sweet! 10's of fA to 1 uA. What diode is that?

George Herold

It's a Vishay PAD5, in SOT-23. I think it's actually a jfet spec'd to be used as a diode. It's not very good at higher currents... 1.5 volts drop at 10 mA.

Someone. Win or Phil H, noted that transistors are better diodes than diodes. Looks like fets are also better diodes than diodes.

JOhn

Thanks John, An npn transitor used as a diode is 'ideal' down to the

1nA or perhaps 100 pA range. Lower currents than that and it becomes 'non-ideal' See Sze "Physics of Semi-devices" Chapeter 3.

George Herold

Maybe now that John's built his fA meter, he can test a few transistors & FETs out for us. ;^)

--
    W
  . | ,. w ,   "Some people are alive only because
   \\|/  \\|/     it is illegal to kill them."    Perna condita delenda est
---^----^---------------------------------------------------------------

I did just test an NPN RF transistor as a diode, just for fun. At +-

0.1 volts, I'm seeing ballpark 20 fA currents, pretty much at my resolution limits. At +-0.2 volts, it's well below 1 pA. I'm seeing a beautiful log curve from about 1 pA up into the microamps at least.

I'd somewhere got the impression that RF transistors are leaky, but apparently not. I guess small junctions are just less leaky than big ones.

Anybody can do this!

John

s e s

Interesting! The only thing Sze says about the low current region is the following. "To improve the current characteristics in the low- current region, the trap densities in the depletion region and at the semiconductor surface must be reduced." I guess people have gotten good at this.

I most admit there is something intriguing about currents so small that you can almost count each electron as it goes by. 20fA ~ 1e/10us

George

Soon they will be able to read the serial number on each electron, as well. ;-)

--
You can\'t have a sense of humor, if you have no sense!

Yeah, surfaces are a lot better that they were, and crystal defects ditto.

One thing I'd like to do one day is a science-project level experiment to demonstrate single electron charges, using Mouser type parts.

One idea is to use an eprom cell as the test circuit; that's a fF floating mosfet gate, and they leak electrons per week or some such.

Another is to use a small-geometry mosfet with the gate floating. Pulse it with a blue LED once in a while, to kick off some small number of electrons at each pulse. Digitize drain current and do statistical stuff to demonstrate quantized current changes after each blip.

Or do some electro-mechanical thing, with an AFM cantelever or something.

Of course, Millikan and Fletcher beat us to it by 100 years.

John

to

olts

han

the

omes

rs

"> One thing I'd like to do one day is a science-project level experiment

Dang! And I was going to ask you if you can measure the noise level in your fA currents. i sub noise =3D root(2*e*I dc) (amps per root Hertz) As I'm sure you know. Measure the shot noise and you measure the charge of the electron. It's only recently that I've understood that if electrons had twice as much charge. They would look like delta functions with twice the height, coming at one half the average rate. (to give equal currents) And this would give twice the noise*.

So, I'm being paid to do just that, making an instrument to measure resistor Johnson noise (and measure either k or T) and shot noise, (measure e). You can see our stuff at

I've measured the shot noise in Ibe of transistors and photodiodes, illuminated by light bulbs, down to the nA level. But you're measuring currents more than four orders of magnitude below that! Yeah I'm interested! The more orders of magnitude I can put on a graph the better it is.

I think I like the mosfet and blue LED idea. You propose to measure the channel conductance, add a bit of light to excite some charge on the gate, then measure conductance again. Looking for quantized steps in the conductance. It should be easy to control the pulse width/ height going to the LED and control the number of photons. But I have no idea how much the conductance changes with the addition of one electron. (of course my understanding of your idea may not be correct.)

I'm not sure who Fletcher was, but Schottky was the man who measured shot noise first... also almost 100 years ago.

George Herold

*noise measured in units of amps or volts squared per hertz.

e:

'd to

volts

than

to the

ecomes

tors

a

us

e

text -

OK I found Fletcher, sounds like Millikan may have been an A-hole, I'd rather work with average good guys, than brilliant A-holes. of course the best are brilliant good guys!

George Herold

Sort of theoretical. I want to see real steps.

It's only recently that I've understood

My setup, using an LMC6001 opamp, has a few fA of low-frequency noise, probably opamp noise and Johnson noise in the gohm resistors. So that's pretty surely burying any shot noise. Just had too much food and beer to do the math reliably.

I tried a voltage divider, 1G and 5G ohm resistors in series, just to check my box. The voltage drop ratio was 5.002, pretty good. I don't know if these gohm-range resistors have shot noise or not. Metal films don't, but high-value thickfilms are claimed to have shot noise.

Hmmm, just got some samples of some 100 G and 1 Tohm surface-mount resistors from IMS. Maybe I'll try to measure one.

Wow, very cool stuff. The NMR is impressive. Gonna do an MRI imager?

That's about it.

Let's see... with, say, a 3 pF gate capacitance, 1e changes the gate voltage by about 50 nV. That might change drain current by a few nanoamps. The numbers aren't totally absurd.

For this to work, you'd prefer to flash the led a lot more often than the average electron leakage rate, so the noise situation wouldn't be awful. Digitize drain current over a bunch of shots over, say, 1 millisecond; flash; repeat. Then see if you can discern steps in the delta-I measurements.

What we really need is a lower-capacitance fet; hence the eprom idea.

John

te:

c'd to

5 volts

s than

to the

becomes

stors

- a g 0us

re

=93My setup, using an LMC6001 opamp, has a few fA of low-frequency noise, probably opamp noise and Johnson noise in the gohm resistors. So that's pretty surely burying any shot noise. Just had too much food and beer to do the math reliably.=94

Yeah the 1 G ohm is going to give you 4 fA/rtHz of current noise. I always think of measuring shot noise in a TIA configuration, with the big resistor as feed back. Then you need a big enough R so that the voltage drop across it is more than 50mV, to get above the Johnson noise. In your configuration if you have say 1pA of current flowing through the diode, there=92s only 1 mV of voltage drop across your 1G ohm resistor and as you say the resistor Johnson noise kills you.

"Wow, very cool stuff. The NMR is impressive. Gonna do an MRI imager?"

I can take no credit for the current incarnation of the pulsed NMR. (Except for the gradient coils and thermal stability of the magnet.) The hard electronics was done by Norman Jarosic in his spare time. He also did a lot of the electronics for WMAP.

We can do some simple imaging in the earth=92s field NMR. Here the sample size is a half liter bottle of water. There is a segmented sample vial and you can put water in some segments and leave it out of others, and then see splitting in the FFT when a field gradient is applied. One cool part of the earths field NMR is that the precession frequency is near 2kHz. We add a speaker to the output and you can hear the FID decay.

=93Let's see... with, say, a 3 pF gate capacitance, 1e changes the gate voltage by about 50 nV. That might change drain current by a few nanoamps. The numbers aren't totally absurd.=94

OK I=92ll have to think about that. Do you just want to show that charge is quantized (seeing steps) or do you think you could get a real number out, (measure the size of the steps)?

Speaking of quantized steps, there=92s a cool quantum demo that I hope some day we might sell. You take two gold wires, bias them with a battery and measure the current that flows through them when they touch. Now you=92ve got to play around a bit and get them to =91just touch=92 as you bang on the table that they are sitting on. When the wires are just coming out of or into contact they form a quantum point contact. Capturing the transient behavior on a storage =91scope you sometimes see a series of steps, the step height is related to e^2/h, the fundamental unit of conductance. 1/12.7 k Ohms (or something like that.)

George Herold

to

volts

than

There must be fun things you can do with flexure-type AFM tips. They taper down to atomic diameters and are cheap. Some of the scanning atomic-force microscopes bounce a pointer-quality laser off the shiny top surface into a pair of photodiodes. Just atomic VanDerWaals forces on the tip cause enough deflection to easily measure.

I suspect you could see individual electrons lost from a charged AFM tip. The capacitance of the tip to the universe must be a fraction of a pF... they are pretty small. I have a bunch around here that I've been meaning to play with. Thing is, I have a bigger incentive to play with the things that make money, as opposed to the things that are purely fun.

Which means I should be heading off to work.

John

Yeah, you've got to make something that someone else wants to buy. Building instruments for students, it's much too easy for me to be distracted by all the cool stuff you can do. But I've got to keep the basics in mind. With the current noise project the first goal is to get them to understand the units. Then we can measure some stuff.

Walking out with the dogs tonight I was thinking about AFM's. I don't know how they drive these things. But maybe I could crash one plated with gold into something else and make some quantum contacts. The last idea we had was to use a piezo and two cylinders at right angles wrapped with gold wire.... There are still some other alignment issues with this idea and it hasn't gone anywhere. I'll keep AFM's in mind, where do we get them cheap?

George Herold

Lots of people sell AFM tips, like...

for something like $10, as I recall.

You could have all sorts of fun with a 10 nm tip. You could certainly make a simple field ion microscope and image the individual atoms of the tip.

With some piezo (or mechanical? magnetostrictive?) actuators for the fine approach, you can certainly do tunneling, and maybe your quantum conduction stuff, too.

Lots of toys around.

John

Stephen Woodward published a Design Idea, I think in EDN, for a circuit used to make STM tips. The article describes the procedure too.

Ah, here it is:

ISTR a mention in SED ages ago that 32kHz watch-crystals made decent AFM actuators. Quartz tuning forks, at any rate, should be orders of magnitude more stable than most cheap alternatives I can think of.

Cheers, James Arthur

t n g

ed

f

and

te

e

st

he

t

ou

ike

e t

Thanks James, I can buy these for $20 in small quantities which just sounds easier. (from John's link in previous post).

As a grad student I etched sharp tips on tungsten wire by a similar method. Tips were used to make ions in liquid helium.

George Herold

Is your resistor really that clean?