Title: QUANTUM COMPUTATION
1QUANTUM COMPUTATION
- The History and the State-of-Art
- Part 1
2Plan
- 1. Conventional and Quantum Logic
- 2. Quantum algorithms
- 3. Quantum Error Correction
- 4. Physical implementation of quantum computation
- 5. Adiabatic quantum computer
- 6. Quantum Simulator
31. Conventional and Quantum Logic
4Turing, 1936 Computer Science
- All computers are equivalent, and could all be
simulated by each other. - A problem is either computable or not, regardless
of what computer we use - Assumption
- Computer operation is based on conventional
logic
5 - Conventional (classical) computer
- Operates with the digital units bits
- Number in binary notation
- N a5a4a3a2a1a0
- ak 0 or 1
- N a020 a121
6- Physical implementation of a bit
- Voltage of a capacitor
- Magnetic moment of a domain
7Classical Computation
- Change of the values of bits
- Logic gates
- Logic gates are based on the conventional logic
- Negation (NOT)
- NOT (True) False
- NOT (False) True
8Classical logic gates
Operation of N-gate
Final state of bit
Initial state of bit
0
1
0
1
92-Bit Logic Gates
- (True) AND (True) True
- Any other combination with AND False
- AND gate
- 00?0
- 01?0
- 10?0
- 11?1
- AND and NOT are universal for classical
computation
10Landauer, 1961 Logic and Thermodynamics
- We use irreversible two-bit gates
- ? Conventional logic is
- thermodynamic logic
- Shannon entropy
- S -kB?plnp
11Classical Computer is a Heat Generator
- In a single computational step
-
- Entropy S of the available information decreases
by kBln2 - Entropy of the environment ?S gt kBln2
- ? Classical computer generates heat Q gt kBT
ln2 - ? Classical computer consumes energy E gt
kBT ln2
12Classical Computer vs Heat Engine
- If a classical computer were built before the
heat engine - the 2nd law of thermodynamics would be formulated
as - A computer based on a classical logic cannot
operate without generation of heat
13Bennett, 1973 Reversible Logic
- A computer can operate using a novel
- reversible logic!
- Toffoli, 1980
- Universal Control-Control-Not 3-bit gate
- 111?110
- 110? 111
- All other combinations do not change
14Benioff, 1980 Quantum Computer (QC)
- Computer based on the non-dissipative classical
physics is possible - Advantage compare to the conventional computer
no basic computational heat generation - (Real heat generation is much greater)
- Computer based on the quantum physics is possible
15- Bit stationary state of an atom or ion
- Advantage
- Huge possible information density
- 1022 bits/cm3
16Deutsch, 1985 Superpositional States
- The most important power of QC
- Using superpositional states
- C00gt C11gt
- Quantum measurement
- C02 probability to collapse
to the state 0gt - C12 probability to collapse
to the state 1gt
17Quantum Logic
- Quantum bit (qubit)
- Superposition of true and false
- A Quantum logic gate must change all the states
of superposition simultaneously
18Operations with Qubits
- Examples
- One qubit
- Initially the ground state 0gt
- One-qubit quantum gate can create the state
- (0gt 1gt)/v2
- Physical implementation
- electromagnetic (EM) pulse
19- Two qubits
- Two one-qubit gates can create the state
- (00gt 01gt 10gt 11gt)/2
- Three qubits
- Three one-qubit gates can create the state
- (000gt 001gt 111gt)/v8
20Loading Numbers
- We can create superposition of all possible
numbers! - L qubits we can load 2L numbers at a time using
L qubits and L one-qubit gates!
21Quantum Logic Gates
- Quantum logic gate - any unitary transformation
(transformation, which does not change ?Ck2
1) - One qubit gate (one qubit rotation) changes the
state of a single atom -
22- Example
- One-qubit Hadamard gate H
- H0gt (0gt 1gt)/v2
- H1gt (0gt - 1gt)/v2
23Hadamard (H) Gate
- 1) H creates the superposition of all possible
numbers from the ground state. - 2) H also implements the simplest quantum
interference - H(0gt 1gt)/v2
- 0gt 1gt)/v2 (0gt - 1gt)/v2/v2
- 0gt
- We got one number from two numbers!
24Universal Quantum Computations Using Quantum
Logic Gates
- Barenco et al., 1995
- Theorem
- For universal QC we need one-qubit rotations and
a two-qubit gate. - The most popular two-qubit gate is the
Control-Not (CN) gate
25Control Not (CN-gate)
Operation of CN-gate
At least two qubits are required
0
0
0
0
Control qubit c
1
0
0
1
Target qubit t
1
0
1
1
0
1
1
1
26An Implementation of the CN Gate
- Interacting atoms.
- The resonant frequency of one atom depends on the
state of the other atom - Frequency is ? if the neighbors state is 0gt
- Frequency is ? if the neighbors state is 1gt
- Using electromagnetic pulse of frequency ? we
drive the atom only if the neighbors state is
1gt!
27Two Registers
- Using two registers we can load 2L values of a
function - Example
- 2 qubits are in the x-register, 1 qubit is in the
y-register
28- x,ygt 00,0gt
- H1H200,0gt
- (00,0gt 01,0gt 10,0gt 11,0gt)/2
- Now CN23 gate changes the value of y
- (2-control qubit, 3-target qubit)
- (00,0gt 01,1gt 10,0gt 11,1gt)/2
- Periodic function y(x)!
- It is an example of the entangled state
29Computing a Function
- In general
- We can load 2L values of a function with a
polynomial number of gates! - How to use this computational power?
30Quantum Interference
- Use quantum interference.
- Eliminate all the numbers except for the desired
ones - Make quantum measurement
- Get the desired parameter of a function!
312. Quantum Algorithms
32Problems for a Classical Computer
- Input L bits
-
- Classical computers can solve
- Polynomial time (P) problems
- number of computational steps increases as a
polynomial function Ln - (typically n 3).
33Polynomial time (P) problems
- Addition, multiplication, solution of an ordinary
differential equation are the - P problems
- Tractable problems
- Efficient algorithm
34Problems for QC
- Non-deterministic polynomial time problems (NP)
- Input L bits (qubits)
- The number of computational steps needed to
verify a solution increases as a polynomial
function Ln. - Number of computational steps needed to solve the
problem increases typically as 2L (exponentially)
35Prime Factorization
- Example 29083 127 x 229
- Gauss
- The dignity of science demands that every aid to
the solution of such a celebrated problem be
zealously cultivated.
36- Statement There is no efficient classical
algorithm for prime factorization - (It is not proved!)
- Since 1970th the main cryptosystem (RSA) is based
on this statement. - Bank cards, diplomatic secret codes are based on
this statement.
37Financial Security
- No one can fake bank cards
- A computer chooses at random two prime factors N1
and N2, - multiply them,
- assign the product N N1N2 to your account
(visible for banks employees), - encode your bank card with one of the factors,
- and erase both factors.
38Prime Factorization on a Classical Computer
- N - 200-digit number
- The best supercomputer
- IBM Blue Gene 1015 operations/second
- If we try about 10100 numbers
- It would take about 1085 s, more than 1077 years
- The best classical algorithm takes
- about 2x109 years
- It is still inefficient
39Shor, 1994 The First Practical Quantum Algorithm
- 1. Create superposition of all possible numbers
in the x-register - 2. Compute f(x) ax mod(N) in the y-register
- (a-is any co-prime to N)
40- 3. Use Shors gate (discrete Fourier transform
DFT) to eliminate all the numbers in the
x-register except for multiples of (2L /T). - (L - number of qubits in the x-register)
- 4. Make quantum measurement on the x-register.
- 5. Repeat procedure to get multiples of (2L/T)
and derive the value of T. - Find a factor GCD (N, aT/2 1)
41- The Shors algorithm is efficient!
- Polynomial number of computational steps (less
than L3) - A few hours instead of 2x109 years!
42Complete NP Problems
- Is there is a hope to find efficient algorithms
for other NP problems? - Cook, Karp, and Levin (1970th)
- There exist complete NP problems
- with the following properties
- an efficient algorithm for a complete NP problem
can be used to construct an efficient algorithm
for any other NP problem!
43- Examples of a complete NP problem
- Given the dimensions of boxes how to pack them
into the trunk? - Given a finite set of integers. Whether any
subset sums to zero?
44- There is no efficient classical algorithm.
- No one proved that it does not exist
- Clay Math Institute in Massachusetts
- 1,000,000.00 reward for the algorithm
- What about an efficient quantum algorithm?
45Grover, 1996
- Quantum speed-up for a very large class of the
NP problem - vN steps instead of N/2 steps
- Example
- We need 1026 operations (1011/s)
- More than 1000 years
46- Using Grovers algorithm
- 1013 operations
- Less than 1 second!
- Impressive but still inefficient!
47How Grovers Algorithm Works
- 1. Create superposition of all possible numbers
in the x-register - 2. In the y-register enter y 1 for x x,
where x is the problems solution (it is easy
to check if x is a solution) and y -1 for all
other x.
48- 3. Use Grovers gate to increase the amplitude
C for the state ltxx,y-1. - 4) Make a quantum measurement to find x.
49- Still exponential number of steps!
- Is it possible to find an efficient quantum
algorithm for a complete NP problem? - Nobody could find it.
- No one proved that it is impossible
50New Physical Principle
- Aaronson, 2008
- A physical computer, which can efficiently solve
a complete NP problem is impossible. - Informational principle
51Physical applications of the Informational
Principle
- Example non-linear correction to the Schrodinger
equation non-linear quantum mechanics - Abrams and Lloyd,1998
- If a non-linear term is added to the Schrodinger
equation then a quantum computer can solve
efficiently a complete NP problem
523. Quantum Error Correction
53Error Correction in a Classical Computer
- Error correction in a classical compute -
redundancy - Take two additional (ancillary) bits,
- All three bits are supposed to have the same
value - Check the values of all the bits
- If one of the bits has a value different from two
other bits change its value
54- How to correct the amplitudes Ck for qubits?
- Classical redundancy cannot be used!
55Steane and Shor, 1995 Quantum Error Correction
- The most complicated quantum algorithm!
- Example
- (C00gt C11gt)00gt C0000gt C1100gt
- Assumption Error occurs independently on
different qubits
56Procedure for Quantum Error Correction
- 1. Use a special sequence of quantum gates to the
three qubits - 2. Measure the state of the ancillary qubits.
- 3. Depending on the measurement result apply a
sequence of quantum gates to the main qubit. - As a result the main qubit returns to the initial
state (C00gt C11gt) - with accuracy to a non-significant phase
constant exp(ikt)
57Collective Decoherence
- What happens if the errors on different qubits
are correlated? - Collective decoherence
- Example
- Thermal electromagnetic wave with the wavelength
longer than the distance between the qubits
58Super-decoherence and sub-decoherence
- Collective decoherence depends on the type of
states - For example, in some cases for L atoms
- Difference ? between the integrated values in a
superposition is about L super-decoherence
quantum superposition quickly destroys - ? ltlt L sub-decoherence
-
59- Example with two atoms
- (00gt 11gt)/v2 super-decoherence
- ? 2 0 2
- (01gt 10gt)/v2 sub-decoherence
- ? 1 1 0
60Palma, Suominen, and Ekert, 1996 Logical Qubits
- Two physical qubits one logical qubit
- 0gt ? 01gt
- 1gt ? 10gt
- (00gt 11gt)/v2 ? (0101gt 1010gt)/v2
- Sub-decoherence instead of super-decoherence!
- Decoherence-free subspace