Download Document

Quantum Computers – Is the
Future Here?
Tal Mor –
CS.Technion
QIPA
Dec. 2015
128 ?? [ 2011 ; sold to LM ]
D-Wave Two :512 ?? [ 2012 ; sold to NASA + Google ]
D-Wave Three: 2,048 ?? [ 2016 ?? ]
Content
• Quantum information and computation –
what for?
• Quantum Bits and Algorithms
• Implementations – Current Status
• “Semi-Quantum” Computing
• Conclusions
Quantum Information – what for?
• First, quantum computers can crack some of the
strongest cryptographic systems (e.g. RSA)
• Second, they might be useful for various other things
as well (simulating quantum systems etc.)
• Quantum cryptography provides new solutions to
some cryptographic problems
• Quantum cryptography may become useful ALSO if
(new) classical algorithms will crack RSA
• Quantum Teleportation and quantum
ECC can enlarge distance for secure
quantum communication
• Satellite quantum communication
CREDIT: Science/AAAS
Quantum Computers – what for?
• Quantum computers can crack RSA because they
can factorize large numbers of n digits in polynomial
time!
O(n2 log n)
• A “classical computer will have to work “subexponenital time”
O(exp[(n log n)1/3])
CREDIT: Science/AAAS
Quantum Computers – what for? (2)
• Quantum computers might be useful for various other
things as well….. Mainly - simulating quantum
systems:
– Fully understanding the complicated electronic
structures of molecules and molecular systems
– Predicting reaction properties and dynamics
– Designing well controlled state preparation
– Analyzing protein folding
– Understanding photosynthetic systems
– HHL. Etc. Etc. Etc.
• The HOPE is to have advantage
already with 30-100 qubits
CREDIT: Science/AAAS
Quantum Computers – what for? (3)
• Quantum algorithms applied onto small “quantum
computers” might be useful for various QUANTUM
TASKS….. Mainly - manipulating quantum
systems:
– Algorithmic cooling of spins, for improving
MRI/MRS/NMR/ESR is one of my team’s goals.
– We collaborate on this topic with the one of the largest
quantum computing centers – in Waterloo.Canada.
– As said before: quantum ECC can
enlarge distance for secure
quantum communication
CREDIT: Science/AAAS
D-Wave collaborations
(Wikipedia)
In 2011 ,Lockheed Martin signed a
contract with D-Wave Systems to
realize the benefits based upon a
quantum annealing processor applied
to some of Lockheed's most challenging
computation problems. The contract
includes the purchase of a “128 qubit
Quantum Computing System”.
In 2013, a “512 qubit system” was
sold to Google and NASA.
D-WAVE: Superconducting flux qubit
MW Johnson et al. Nature 473, 194-198 (May 2011)
However, their “qubits” are highly limited. Similar Technology
with less limited qubits reached 3-5 qubits, no more!
So what is the TRUTH??
The Qubit
In addition to the regular values {0,1} of a bit, and
a probability distribution over these values, the
Quantum bit can also be in a superposition
www.cqed.org/IMG/jpg/compdoublemobilemz.jpg
The Qubit (2)
A superposition state α|0› + β|1›
Intereference (as in waves)
scienceblogs.com
http ://upload.wikimedia.org/wikipedia/commons/2/2c/Two_sources_interference.gif
The Qubit (2)
A superposition state α|0› + β|1›
Intereference (as in waves)
scienceblogs.com
http ://upload.wikimedia.org/wikipedia/commons/2/2c/Two_sources_interference.gif
The Qubit (2)
A superposition state α|0› + β|1›
… with |α|2 + |β|2 = 1
scienceblogs.com
http ://upload.wikimedia.org/wikipedia/commons/2/2c/Two_sources_interference.gif
The Qubit (3)
•
|+› = (1/√2) |0› + (1/√2) |1›
Is a state of a particle in BOTH arms!
• But this is ALSO a state of a particle …
in both arms
|-› = (1/√2) |0› - (1/√2) |1›
The amplitude 1/√2 means a probability of 1/2 to find the
particle if we look for it, in one arm or another
[In general: α|0› + β|1›  P(0)=|α|2 and P(1)= |β|2 ]
The Qubit (4)
• The two arms meet - there is an interference
• This is so due to Linearity of quantum mechanics
• |0› → |+› = (1/√2) |0› + (1/√2) |1›
|1› → |-› = (1/√2) |0› - (1/√2) |1›
• We get
›
›
›
|+ = (1/√2) |0 + (1/√2) |1 →
›
›
› - (1/√2) |1›]
(1/√2) [(1/√2) |0 + (1/√2) |1 ] + (1/√2) [(1/√2) |0
›
= |0
“Constructive/Destructive Interference”
Two Qubits - Entanglement
α|00› + β|11›
brusselsjournal.com
Initialization
(via Hadamard transformation)
• Single qubit: |0›  |+› = [|0›+|1›] / (1/√2)
• Two qubits: |00›  |++›
= (1/2) [|00›+|01›+|10›+|11›]
• For n qubits: |++ … +› = 1/(2n/2) ∑ i=0 |i›
A superposition over all 2n possible values
16
Initialization
(via Hadamard transformation)
• Single qubit: |0›  |+› = [|0›+|1›] / (1/√2)
• Note: the Hadamard transformation is also
responsible for the interference!!!
|+›  |0›
|-›  |1›
17
n Qubits – parallel computing
• Prepare a superposition
over 2n states
• Run your algorithm
in parallel
• Interference enhances the
probability of the desired solution futuredocsblog.com
• Peter Shor factorized large numbers (in principle)
using Shor’s algorithm!
• Several other problems in NP were also solved
• Current quantum architecture reaches 13-14
qubits (NMR, ion trap); far from being practical…
Will quantum computers
factorize large numbers?
• If ‘yes’ – this is a revolution in Computer
Science
• If ‘no’ – this is a revolution in Physics
• So let’s assume it will… but maybe not so
soon!
• Can we predict when?
19
futuredocsblog.com
Implementations
1. Ion trap (qubit is the ground-state vs
excited-state of an electron attached to an
ion; “many” ions in one trap)
2. NMR (qubit is the spin of a nuclei on a
molecule; “many” spins on a molecule)
3. Josephson-Junction qubits (magnetic flux)
4. Optical qubits (photons)
• Etc…
Example – ion trap
Reached 14 (8) qubits •
Nobel Prize and Wolf Prize •
sciencedaily.com
NIST
Example - NMR
Reached 13 qubits •
Scalability problem •
Resolved via Algorithmic Cooling* •
tudelft.nl
robert.nowotniak.com
Examples 3+4
Josephson Junctions (8-9 qubits, DWAVE 8**?) •
Q. Optics (6-7 qubits) •
Sufficient for some ECC •
The Australian Centre of Excellence for
Quantum Computation and Communication Technology
Current status of fullyquantum computing
• Despite the Nobel prize – we have no clue
when ion traps will reach 25 qubits
• Despite of 20M $ DWAVE computers
already sold – we have no clue if JJ qubits
are of any good; We do know (Shin, Smith
Smolin, Vazirani; 2014) that there is
probably no reason to believe that the
DWAVE model is quantum**.
Current status of fullyquantum computing
• Optmism – due to Moore’s Law for VLSI
• Optimism – Semi QC
Limited QC Models:
Semi-quantum computing
• D-Wave’s AQC[???]; Other Q. Sim.?
• One Clean Qubit * (closely related to NMR)
• Linear Optics (closely related to Q. Optics)
Three Extremely Important Questions:
• What algorithms can the limited models run?
[OCQ – Trace estimation; LO – boson sampling]
• What kind of Quantumness/Entanglement is there?
• Do they scale much easier/better than full QC?
(no) Conclusions
• Zero conclusions about the future of full QC
• Some optimism about semi-quantum
computing
• Many more questions than answers, both
theoretically and experimentally
Thanks
Appendix – DJ Algorithm
• Initialization (1 slide)
• Parallel computing (1 slide)
• Deutsch’s Algorithm (6 slides):
1. Deutsch Problem (1 slide)
2. Classical solution (1 line)
3. Quantum solution (the remaining 5 slides)
28
Initialization
( via Hadamrd Transformation)
• Single qubit: |0›  |+› = [|0›+|1›] / (1/√2)
• Two qubits: |00›  |++›
= (1/2) [|00›+|01›+|10›+|11›]
• For n qubits: |++ … +› = 1/(2n/2) ∑ i=0 |i›
A superposition over all 2n possible values
29
Parallel Computing
The quantum computer computes in parallel
• If we know to classically apply U onto
|00…0000›, onto |00…0001› etc,
efficiently, we also immediately know how
to apply it to that superposition
|++…++++›
• We get parallel computing over 2n values!
• Interference yields the desired outcome!!
30
The first quantum algorithm
(Deutsch, 1989)
• We are given a black box calculating f(x),
where x is a single bit, 0 or 1, and f(x) is also
a single bit
• We do not care what is f(0) and f(1), we only
want to know if f(0)=f(1) [ is f constant? ]
• Each call to the black box costs 1 Billion
dollars, if we guess correctly the answer, we
win 1.5 Billion dollars, if we fail we lose 2
Billions dollars! Can we become rich?
31
The first quantum algorithm
(Deutsch, 1989) (2)
• Classically we lose of course….
• With a quantum BOX – we may become rich!
• We do not care for f(0) and f(1), we only need
[f(0)==f(1)] ; We calculate this DIRECTLY!!
• We shall use an additional (ancilla) qubit for
keeping the result f(x): |x›|0› → |x›|f(x)›
• We also plan that:
|x›|1› → |x›|f(x)`›
Did we become rich?
32
The first quantum algorithm
(Deutsch, 1989) (3)
•
•
•
•
•
•
We plan the transformation: |x›|0› → |x›|f(x)›
We plan also:
|x›|1› → |x›|f(x)`›
We prepare our ancilla: |-› = [|0›-|1›] / (1/√2)
We apply |x›|-› → |x›[|f(x)› - |f(x)`›] / (1/√2)
For x=0 we get: |0›[|f(0)› - |f(0)`›] / (1/√2)
For x=1 we get: |1›[|f(1)› - |f(1)`›] / (1/√2)
Did we become rich?
33
The first quantum algorithm
(Deutsch, 1989) (4)
• For x=0 we get: |0›[|f(0)› - |f(0)`›] / (1/√2)
• For x=1 we get: |1›[|f(1)› - |f(1)`›] / (1/√2)
• But…. We prepare a superposition for x as
well: [|0›+|1›] / (1/√2)…
• And we get
{ |0›[|f(0)› - |f(0)`›] + |1›[|f(1)› - |f(1)`›] } / 2
Did we become rich?
34
The first quantum algorithm
(Deutsch, 1989) (5)
{ |0›[|f(0)› - |f(0)`›] + |1›[|f(1)› - |f(1)`›] } / 2
• Case 1: f(0)=f(1)=b [ f is constant ]:
{ |0›[|b› - |b`›] + |1›[|b› - |b`›] } / 2
{ [|0› + |1›] |b› - [|0› + |1›] |b’› } / 2
• Case 2: f(0)=c ; f(1)=c’:
{ |0›[|c› - |c`›] + |1›[|c’› - |c›] } / 2
{ [|0› - |1›] |c› - [|0› - |1›] |c’› } / 2
35
Did we become rich?
The first quantum algorithm
(Deutsch, 1989) (6)
Let’s apply Hadamard transformation onto the left
qubit, and then measure it:
• Case 1: f(0)=f(1)=b [ f is constant ]:
{ [|0› + |1›] |b› - [|0› + |1›] |b’› } / 2
 [|0›|b› - |0› |b’›] / √2 = |0›[|b› - |b’›] / √2
• Case 2: f(0)=c ; f(1)=c’:
{ [|0› - |1›] |c› - [|0› - |1›] |c’› } / 2
 [|1›|c› - |1› |c’›] / √2 = |1›[|c› - |c’›] / √2
We became rich!
36