## Fast parallel circuits for the quantum Fourier transform (2000)

### Cached

### Download Links

Venue: | PROCEEDINGS 41ST ANNUAL SYMPOSIUM ON FOUNDATIONS OF COMPUTER SCIENCE (FOCS’00) |

Citations: | 51 - 2 self |

### BibTeX

@INPROCEEDINGS{Cleve00fastparallel,

author = {Richard Cleve and John Watrous},

title = { Fast parallel circuits for the quantum Fourier transform},

booktitle = {PROCEEDINGS 41ST ANNUAL SYMPOSIUM ON FOUNDATIONS OF COMPUTER SCIENCE (FOCS’00)},

year = {2000},

pages = {526--536},

publisher = {}

}

### Years of Citing Articles

### OpenURL

### Abstract

We give new bounds on the circuit complexity of the quantum Fourier transform (QFT). We give an upper bound of O(log n + log log(1/ε)) on the circuit depth for computing an approximation of the QFT with respect to the modulus 2 n with error bounded by ε. Thus, even for exponentially small error, our circuits have depth O(log n). The best previous depth bound was O(n), even for approximations with constant error. Moreover, our circuits have size O(n log(n/ε)). We also give an upper bound of O(n(log n) 2 log log n) on the circuit size of the exact QFT modulo 2 n, for which the best previous bound was O(n 2). As an application of the above depth bound, we show that Shor’s factoring algorithm may be based on quantum circuits with depth only O(log n) and polynomial-size, in combination with classical polynomial-time pre- and post-processing. In the language of computational complexity, this implies that factoring is in the complexity class ZPP BQNC, where BQNC is the class of problems computable with bounded-error probability by quantum circuits with polylogarithmic depth and polynomial size. Finally, we prove an Ω(log n) lower bound on the depth complexity of approximations of the

### Citations

882 | Polynomial-time algorithms for prime factorization and discrete logarithms on a quantum computer
- Shor
- 1997
(Show Context)
Citation Context ...complexity of the quantum Fourier transform (QFT). The quantum Fourier transform is the key quantum operation at the heart of Shor’s quantum algorithms for factoring and computing discrete logarithms =-=[35]-=- and the known extensions and variants of these algorithms (see, e.g., Kitaev [25], Boneh and Lipton [7], Grigoriev [20], and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform... |

847 | A fast quantum mechanical algorithm for database search
- Grover
- 1996
(Show Context)
Citation Context ... [25], Boneh and Lipton [7], Grigoriev [20], and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique =-=[21]-=-, due to Brassard, Høyer, and Tapp [8] and Mosca [29]. In order to discuss the quantum Fourier transform in greater detail we recall the discrete Fourier transform (DFT); for a given dimension m the d... |

807 | Algorithms for quantum computation: Discrete logarithms and factoring
- Shor
- 1994
(Show Context)
Citation Context ... (2) For certain values of m there are very efficient quantum algorithms for the QFT. The fact that the quantum circuit size can be polynomial in log m for some values of m was first observed by Shor =-=[34]-=- and is of critical importance in his polynomial-time algorithms for prime factorization and discrete logarithms. Shor’s original method may be described as a “mixed-radix” method, and is discussed fu... |

663 |
Tukey, “An algorithm for the machine calculation of complex Fourier series
- Cooley, W
- 1965
(Show Context)
Citation Context ...ransform (FFT), which is an algorithm that computes the DFT with O(m log m) arithmetic operations, as opposed to the obvious O(m 2 ) method. The FFT algorithm was proposed by Cooley and Tukey in 1965 =-=[13]-=-, though its origins can be traced back to Gauss in 1866 [17]. The FFT plays an important role in digital signal processing, and it has been suggested [16] as a contender for the second most important... |

459 |
Logical reversibility of computation
- Bennett
- 1973
(Show Context)
Citation Context ...s that it is sufficient to implement each one efficiently by a classical (nonreversible) circuit, and then combine these to produce a reversible circuit by standard techniques in reversible computing =-=[5]-=-. The telescoping subtraction clearly consists of k − 1 subtractions that can be performed in parallel, so the nonreversible size and depth bounds are O(kn) and O(log n) respectively. The prefix addit... |

429 |
zur Gathen and
- von
- 1991
(Show Context)
Citation Context ...thm was proposed by Cooley and Tukey in 1965 [13], though its origins can be traced back to Gauss in 1866 [17]. The FFT plays an important role in digital signal processing, and it has been suggested =-=[16]-=- as a contender for the second most important nontrivial algorithm in practice, after fast sorting. The DFT (and the FFT algorithm) generalize to certain algebraic structures, such as rings containing... |

313 | Strengths and weaknesses of quantum computing
- Bennett, Bernstein, et al.
- 1997
(Show Context)
Citation Context ...ψx〉 has size O(n log(n/ε)) and depth O(log n+log log(n/ε)) = O(log n+log log(1/ε)). It remains to argue that the circuit operates with error O(ε). This follows from standard results based on ideas in =-=[6]-=- about converting quantum circuits that perform measurements and produce classical information with small error probability into unitary operations (without measurements) that can operate on data in s... |

296 |
Approximate formulas for some functions of prime numbers
- Rosser, Schoenfeld
- 1962
(Show Context)
Citation Context ...f bits of m, and each mj has length logarithmic in the length of m. Taking mj to be the j th prime results in a sufficiently dense collection of moduli m for factoring [34] (see Rosser and Schoenfeld =-=[32]-=- for explicit bounds and a detailed analysis of such bounds). Although stated somewhat differently by Shor, the mixed radix QFT method may be described as follows: 1. For j = 1,... ,k define fj = m mj... |

272 | Parallel prefix computation
- Ladner, Fischer
- 1980
(Show Context)
Citation Context ...tively. The prefix addition is a little more complicated. It relies on a combination of well-known tools in classical circuit design. One of them is the following general result of Ladner and Fischer =-=[26]-=- about parallel prefix computations. Theorem 5 (Ladner and Fischer) For any associative binary operation ◦, the mapping (x1, x2, ... , xk) ↦→ (x1, x1 ◦ x2, ... , x1 ◦ x2 ◦ · · · ◦ xk) (8) can be compu... |

211 |
A taxonomy of problems with fast parallel algorithms
- Cook
- 1985
(Show Context)
Citation Context ...plexity class BQNC as all computational problems that can be solved by quantum circuits with poly-logarithmic depth and polynomial size—a reasonably natural extension of previous notation (see, e.g., =-=[11, 28]-=-)—then Theorem 3 implies that the factoring problem is in ZPP BQNC . Finally, we consider the minimum depth required for approximating the QFT. It is fairly easy to show that computing the QFT exactly... |

200 | Elementary gates for quantum computation
- Barenco, Bennett, et al.
- 1995
(Show Context)
Citation Context ...ess. Originally Shor suggests implementing each of these operations by circuits of size mj (not log mj), since any quantum operation can be computed by circuits with exponential-size quantum circuits =-=[3]-=-. This results in a linear-depth circuit overall, although the circuit will be exact. However, we may try to compute each Fmj more efficiently. There are a few possibilities for how to do this, all (a... |

192 | Fault-Tolerant Quantum Computation With Constant Error Rate,” ArXiv e-prints - Aharonov, Ben-Or - 1999 |

147 | Quantum measurements and the abelian stabilizer problem
- Kitaev
(Show Context)
Citation Context ...s the key quantum operation at the heart of Shor’s quantum algorithms for factoring and computing discrete logarithms [35] and the known extensions and variants of these algorithms (see, e.g., Kitaev =-=[25]-=-, Boneh and Lipton [7], Grigoriev [20], and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21],... |

143 | Mosca: Quantum algorithms revisited
- Cleve, Ekert, et al.
- 1998
(Show Context)
Citation Context ...d computing discrete logarithms [35] and the known extensions and variants of these algorithms (see, e.g., Kitaev [25], Boneh and Lipton [7], Grigoriev [20], and Cleve, Ekert, Macchiavello, and Mosca =-=[10]-=-). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21], due to Brassard, Høyer, and Tapp [8] and Mosca [29]. In order to discuss the quantum ... |

126 | Modern Cryptography, Probabilistic Proofs and Pseudorandomness
- Goldreich
- 1999
(Show Context)
Citation Context ...ed to implement these computations as reversible circuits. We now describe each of the above steps in more detail. Let us first recall the following fact from probability theory (see, e.g., Goldreich =-=[18]-=-). If X1,... ,Xt are independent Bernoulli trials with probability pX of success and Y1,... ,Yt are independent Bernoulli trials with pY of success, where pX < pY , then [ t∑ t∑ ] Pr Xi ≥ < 2e −(pY −p... |

116 |
Log depth circuits for division and related problems
- Beame, A, et al.
- 1986
(Show Context)
Citation Context ...g n) 2 ) and size O(n3 ). There are alternative methods for performing iterated multiplication achieving various combinations of depth and size. In particular, it was proved by Beame, Cook and Hoover =-=[4]-=- that a product such as we have above can be computed by O(log n) depth boolean circuits of size O(n5 (log n) 2 ). While O(n5 log n) qubits may seem a high price to pay in order to save a factor of O(... |

101 | Quantum counting
- Brassard, Høyer, et al.
(Show Context)
Citation Context ...[20], and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21], due to Brassard, Høyer, and Tapp =-=[8]-=- and Mosca [29]. In order to discuss the quantum Fourier transform in greater detail we recall the discrete Fourier transform (DFT); for a given dimension m the discrete Fourier transform is a linear ... |

88 |
An Approximate Fourier Transform Useful in Quantum Factoring
- Coppersmith
- 1994
(Show Context)
Citation Context ...d is discussed further in Section 7.2. In the particular case where m = 2 n , there exist quantum circuits performing the quantum Fourier transform with O(n 2 ) gates, which was proved by Coppersmith =-=[14]-=- (see also [9]). These circuits are based on a recursive description of the QFT that is analogous to the description of the DFT exploited by the FFT. While in some sense these quantum circuits are exp... |

78 |
Schnelle Multiplikation großer Zahlen
- Schönhage, Strassen
- 1971
(Show Context)
Citation Context ...ntaining primitive m th roots of unity (which can play the role of e 2πi/m in Eq. 1). This more abstract type of FFT is a principal component in Schönhage and Strassen’s fast multiplication algorithm =-=[33]-=-, which can be expressed as circuits of size O(n log n log log n) for multiplying n-bit integers. For more applications—of which there are many—and historical information, see [27, 12, 23]. The quantu... |

60 | Quantum cryptanalysis of hidden linear functions (ex- tended abstract
- Boneh, Lipton
- 1995
(Show Context)
Citation Context ...tion at the heart of Shor’s quantum algorithms for factoring and computing discrete logarithms [35] and the known extensions and variants of these algorithms (see, e.g., Kitaev [25], Boneh and Lipton =-=[7]-=-, Grigoriev [20], and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21], due to Brassard, Høye... |

51 | Rockmore.Generalized FFTs -a survey of some recent results
- Maslen, N
- 1995
(Show Context)
Citation Context ...plication algorithm [33], which can be expressed as circuits of size O(n log n log log n) for multiplying n-bit integers. For more applications—of which there are many—and historical information, see =-=[27, 12, 23]-=-. The quantum Fourier transform (QFT) is a unitary operation that essentially performs the DFT on the amplitude vector of a quantum state—the QFT maps the quantum state ∑m−1 x=0 αx|x〉 to the state ∑m−... |

50 | Distinguishability and Accessible Information in Quantum Theory
- Fuchs
- 1996
(Show Context)
Citation Context ...hat approximates the inverse QFT with precision 1 10 . In this section, since we will need to consider distances between mixed states, we adopt the trace distance as a measure of distance (see, e.g., =-=[15]-=-). The trace distance between two states with respective density operators ρ and σ is given as D(ρ,σ) = 1 2Tr|ρ − σ|, (12) where, for an operator A, |A| = √ A † A. For a pair of pure states |φ〉 and |φ... |

43 | Gauss and the history of the fast fourier transform
- Heideman, Johnson, et al.
- 1984
(Show Context)
Citation Context ...plication algorithm [33], which can be expressed as circuits of size O(n log n log log n) for multiplying n-bit integers. For more applications—of which there are many—and historical information, see =-=[27, 12, 23]-=-. The quantum Fourier transform (QFT) is a unitary operation that essentially performs the DFT on the amplitude vector of a quantum state—the QFT maps the quantum state ∑m−1 x=0 αx|x〉 to the state ∑m−... |

39 |
Limits to Parallel Computation
- Greenlaw, Hoover, et al.
- 1995
(Show Context)
Citation Context ...ng polynomial-size circuits with depth poly-logarithmic in n in the classical setting. Are there logarithmic- or poly-logarithmic-depth quantum circuits for these problems? Greenlaw, Hoover and Ruzzo =-=[19]-=- list several other problems not known to be classically parallelizable, all of which are interesting problems to consider in the quantum setting. Acknowledgments We thank Wayne Eberly for helpful dis... |

32 |
On the algorithmic complexity of discrete functions
- Ofman
- 1963
(Show Context)
Citation Context ...ation of |µ0.xj... x0 〉. (respectively), where x,y ∈ {0,1} n and additions and subtractions are performed as integers modulo 2 n . By appealing to classical results about the complexity of arithmetic =-=[30]-=-, one can construct quantum circuits of size O(n) and depth O(log n) for these operations (using an ancilla of size O(n)). It is straightforward to show that applying a subtraction to the state |ψx〉|ψ... |

28 |
Logarithmic depth circuits for algebraic functions
- Reif
- 1986
(Show Context)
Citation Context ...are not logspace-uniform but rather are polynomial-time uniform; the best known bound on circuit depth for iterated products in the case of logspace uniform circuits is O(log n log log n) due to Reif =-=[31]-=-. 146 Lower bounds Logarithmic-depth lower bounds for exact computations with two-qubit gates are fairly easy to obtain, based on the fact that the state of some output qubit (usually) critically dep... |

25 |
Algorithmic Number Theory, Volume I: Efficient Algorithms
- Bach, Shallit
- 1996
(Show Context)
Citation Context ...fying the promise. for j = 1,... ,n. The following algorithm solves this problem: 1. Define A0 = ( ) ( ) ( ) ( ) 1 0 1 1 0 1 0 0 , A1 = , A2 = , A3 = . 0 1 0 0 1 0 1 1 2. Let xj = Alj Alj−1 · · · Al1 =-=[2,1]-=- for each j, and output x = xn · · · x1. Let us now demonstrate that the algorithm is correct. We note that it is straightforward to show that for a given input l1,... ,ln there is at most one x satis... |

22 | Polynomial simulations of decohered quantum computers
- Aharonov, Ben-Or
- 1996
(Show Context)
Citation Context ..., circuit depth is likely to be particularly relevant in the quantum setting for physical reasons. Perhaps most notably, fault tolerant quantum computation necessarily requires parallelization anyway =-=[1]-=-—under various noise models, error correction must continually be applied in parallel to the qubits of a quantum computer, even when the qubits are doing nothing. In such models, parallelization saves... |

19 | Quantum Fourier sampling simplified
- Hales, Hallgren
- 1999
(Show Context)
Citation Context ...lo m to within accuracy ε for n = ⌈log m⌉. The size of the circuit is polynomial in n + log(1/ε). The method exploits a relation between QFTs with different moduli that was used by Hales and Hallgren =-=[22]-=- in regard to the Fourier Sampling problem (see also Høyer [24] for an extension and simplified proof). The basic components of the technique are as follows: 1. Create a Fourier state with modulus m, ... |

16 |
Nachlass: Theoria interpolationis methodo nova tractata”, Werke, Band 3. Königliche Gesellschaft der Wissenschaften
- Gauss
- 1866
(Show Context)
Citation Context ...ith O(m log m) arithmetic operations, as opposed to the obvious O(m 2 ) method. The FFT algorithm was proposed by Cooley and Tukey in 1965 [13], though its origins can be traced back to Gauss in 1866 =-=[17]-=-. The FFT plays an important role in digital signal processing, and it has been suggested [16] as a contender for the second most important nontrivial algorithm in practice, after fast sorting. The DF... |

8 | Testing the shift–equivalence of polynomials using quantum mechanics
- Grigoriev
- 1996
(Show Context)
Citation Context ...rt of Shor’s quantum algorithms for factoring and computing discrete logarithms [35] and the known extensions and variants of these algorithms (see, e.g., Kitaev [25], Boneh and Lipton [7], Grigoriev =-=[20]-=-, and Cleve, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21], due to Brassard, Høyer, and Tapp [8] ... |

8 |
Quantum searching and counting by eigenvector analysis
- Mosca
- 1998
(Show Context)
Citation Context ...e, Ekert, Macchiavello, and Mosca [10]). The quantum Fourier transform also plays a key role in extensions of Grover’s quantum searching technique [21], due to Brassard, Høyer, and Tapp [8] and Mosca =-=[29]-=-. In order to discuss the quantum Fourier transform in greater detail we recall the discrete Fourier transform (DFT); for a given dimension m the discrete Fourier transform is a linear operator on ∗ E... |

5 |
Quantum Algorithms
- Høyer
- 2000
(Show Context)
Citation Context ...uit is polynomial in n + log(1/ε). The method exploits a relation between QFTs with different moduli that was used by Hales and Hallgren [22] in regard to the Fourier Sampling problem (see also Høyer =-=[24]-=- for an extension and simplified proof). The basic components of the technique are as follows: 1. Create a Fourier state with modulus m, which is the mapping 2. Copy the Fourier state, which is the ma... |

4 |
A note on computing quantum Fourier transforms by quantum programs. Manuscript available at http://www.cpsc.ucalgary.ca/ cleve/papers.html
- Cleve
- 1994
(Show Context)
Citation Context ...further in Section 7.2. In the particular case where m = 2 n , there exist quantum circuits performing the quantum Fourier transform with O(n 2 ) gates, which was proved by Coppersmith [14] (see also =-=[9]-=-). These circuits are based on a recursive description of the QFT that is analogous to the description of the DFT exploited by the FFT. While in some sense these quantum circuits are exponentially fas... |